Method for automatic formation of procedure of generating predicted pixel value, image encoding method, mage decoding method, corresponding device, corresponding programmes and data media storing programmes

FIELD: information technology.

SUBSTANCE: disclosed is use of a parent population which is generated via random formation of a procedure for generating a predicted value, each indicated by a tree structure, and a set of procedures for generating a predicted value is selected as a parent from such a population. The procedure for generating a predicted value is generated as a descendant based on a certain method of development of the tree structure which develops selected procedures for generating a predicted value, where the existing function for generating a predicted value can be a tree end node. The procedure for generating a predicted value, having the best estimate cost, is selected from procedures for generating a predicted value as a parent and a descendant, and overall information content for representing the tree structure and volume of the code, estimated by the predicted pixel value, is used as a cost estimate, and the final procedure for generating a predicted value is formed by repeating the relevant operation.

EFFECT: high encoding efficiency.

28 cl, 14 dwg

 

The technical field

The invention relates to a method of automatic generation procedure of generating predicted pixel values for the automatic generation procedure of generating predicted pixel values that are used to implement high-precision prediction pixel value, and the corresponding device; method of encoding images for efficient encoding of images using the procedure of generating predicted pixel value generated by the above method, and the corresponding device; the method of decoding images for efficient decoding encoded data generated by relevant coding of images, and the corresponding device; programs used to implement the above methods; and computer-readable storage media for storing programs.

This application claims the priority of patent application Japan 2008-275811, filed October 27, 2008, the contents of which are incorporated here by reference.

The level of technology

Usually when encoding images, the value of each pixel of the target encoding is predicted using a previously decoded the same or higher pixels, and the residue prediction is encoded.

According to this method, the coding is of predictive when coding the target pixel (marked "p"), subject to encoding, the prediction value p is generated by using the fact that the previously decoded peripheral pixels (e.g., Inw, In, Ine, and Iw in Fig. 14) have, in General, a high correlation with p, and, in fact, using such peripheral pixels. Further, the predicted value of p is denoted by p'. In the next step forecast error p-p' is subjected to entropy encoding.

For example, bespattering mode in JPEG (see non-Patent document 1) has seven types of predictors, and one selected from them is used for prediction and coding of the pixel value.

In the example referred to as "averaged prediction" as one of the methods in the JPEG predictors, the prediction is done by calculating the average value In and Iw as follows:

x'=(In+Iw)/2 Formula (1)

There are also six other forecasting methods (in addition to the above), which include:

x'=In+Iw-Inw-plane prediction Formula (2)

x'=In forecasting the previous value of the Formula (3)

x'=Inw+(In-Iw)/2 complex prediction Formula (4)

JPEG-LS (see non-Patent document 2), with a higher level of efficiency than JPEG, uses a little more complex forecasting method called "prediction of MED", below.

if Inw≥max (Iw, In) then

x'=min (Iw, In)

else if Inw≤min (Iw, In) then

x'=max (Iw, In)

else

x'=Iw+In-Inw

where max (x, y) - function that returns the x and y values that is larger and mix (x, y) - function that returns the value of x and y, which is smaller.

In addition, a well-known method comprising the job weighted average for the peripheral pixels as the predicted values. According to the simplified method, the weight of each peripheral pixel can be calculated using the least squares method for each image, or you can implement a method of optimization of the coefficients to minimize the relevant amount of code (see non-Patent document 3).

Additionally, although this does not apply to predictive coding, in non-Patent document 4 discloses the optimization of the encoding parameters for encoding images or video using a genetic algorithm (GA), where "template" for generating context that is used to encode binary image is modified using a genetic algorithm, which improves efficiency. Thus, the template is treated as a parameter, and use a fixed encoding procedure.

As a similar method, related to the direction, in non-Patent document 5 discloses the use of a genetic algorithm for the dynamic and the changes are separated form the unit square, encode that relevant increases efficiency. By analogy with the pattern in non-Patent document 4 coding procedure in this case is fixed.

Document technique

Non-patent document

Non-patent document 1: ISO/IEC SC29/WG1 ISO/IEC 10918-1, Digital compression and coding of continuous-tone still images", p. 133, 1993

Non-patent document 2: M. Weinberger, G. Seroussi, and G. Sapiro, "The LOCO-I Lossless Image Compression Algorithm: Principles and Standardization into JPEG-LS", IEEE Trans. Image Processing, Vol. 9, No. 8, pp. 1309-1324, August 2000

Non-patent document 3: Ichiro Matsuda, Nau Ozaki, Yuji Umezu, and Susumu Itoh, "Lossless Coding Using Variable Block-Size Adaptive Prediction Optimized for Each Image", Proceedings of 13th European Signal Processing Conference (EUSIPCO 2005), WedAmPO3, Sep. 2005

Non-patent document 4: Masaharu Tanaka, Hidenori Sakanashi, Masanobu Mizoguchi and Tetsuya Higuchi, "a Bi-level Image Coding for Digital Printing Using Genetic Algorithm", Proceedings of IEICE, D-II, Vol. J83-D-II, No. 5, pp. 1274-1283, May 2000

Non-patent document 5: Koh'ichi Takagi, Atsushi Koike engineering Germany, Shuichi Matsumoto, and Hideo Yamamoto, "the Moving Picture Coding Based on Region Segmentation Using Genetic Algorithm", Proceedings of IEICE, D-II, Vol. J83-D-II, No. 6, pp. 1437-1445, June 2000

The invention

The objective of the invention

As described above, the traditional method of forecasting has flexibility only for optimization of numerical parameters, such as weights for each image, and the forecasting procedure for the determination of the pixel is used to compute predictive, or the formula used for conditional branching is the tsya fixed.

Thus, traditionally, a new forecasting procedure can generate only people manually, by trial and error. Thus, the structure of the corresponding predictor may not be more difficult than to understand people.

In addition, there is no traditional way to re-create the special procedures of the prediction for each input image.

Additionally, when the image processing target image to be processed (i.e. training information or resource)must be manually generated and sustained by the person.

In light of the above circumstances, the present invention is the provision of new techniques to improve the efficiency of encoding and decoding, due to the implementation of automatic computer forming the forecasting process, which is suitably applied to the input image, and to further reduce the relevant amount of code. Here, by analogy with the traditional methods of forecasting, the present invention also uses the previously decoded peripheral pixels to generate the predicted values.

The solution of the problem

<1> structure of the device automatic generation procedure of generating predicted values of the pixel corresponding to the present invention

P is IDE all, consider the structure of the device automatic generation procedure of generating predicted values of the pixel in accordance with the present invention.

The automatic formation procedure of generating predicted values of the pixel in accordance with the present invention, implements the automatic generation procedure of generating predicted values for the prediction value of the target pixel encoding by using the previously decoded pixel. The device has a structure which includes:

(1) the first device that generates the parent population by random generation procedure of generating predicted values, each of which is specified through a tree structure;

(2) a second device which selects a set of procedures for generating the predicted values as parents from the parent population and generates one or more procedures generate predicted values as descendants on the basis of a predetermined method development (or evolution) of a tree structure, which exposes the selected procedure of generating predicted values development (or evolution), where the current function of generating predicted values can be a leaf node of the tree;

(3) t the e device that:

selects the procedure of generating the predicted values with the minimum cost evaluation of the procedures generating the predicted values as the parents and children, where the total information content to represent the tree structure and the amount of code estimated predicted value of the pixel obtained through a tree structure, is used as the cost of the evaluation and the selection procedure of generating the predicted value is the value of the best estimate of the encoding target image coding; and

saves the selected procedure to generate predicted values and one or more other procedures generate predicted values in the parent population; and

(4) a fourth device that manages to repeat the processes carried out by the second and third device until the predetermined condition, and generates a procedure to generate predicted values, with the value of the best estimate, as a result of repetition as the final procedure of generating predicted values.

In the above structure, the second device may establish procedures for the generation of predicted values as descendants on the basis of a predefined way again in the party tree structure, responsible for the development, where the function that prints the coordinates of the pixel in the image can be a leaf node of the tree.

In addition, the first device may generate the parent population so that the current function of generating predicted values was included in the parent population.

The method of automatic generation procedure of generating predicted values of the pixel corresponding to the present invention, implemented through the operations of the above-described processing devices, can also be implemented through computer program. A computer program is stored on an appropriate computer-readable storage medium or through a network, and the invention is installed and executed on the control device, such as CPU, thanks to which the invention is implemented.

In the automatic formation procedure of generating predicted values of the pixel having the structure described above, when the parent population is generated by randomly generate procedures generate predicted values, each of which is specified through a tree structure, the set of procedures for generating the predicted values are selected as parents from the parent population. Then one Il the several procedures of generating predicted values is formed as a descendants on the basis of a predetermined method development tree structure, which puts the development of the selected procedure of generating predicted values. Then select a procedure of generating the predicted values with the minimum cost, where the total content (which can be obtained according to Algorithm 1, described below) to represent the tree structure and the amount of code estimated predicted value of the pixel (which can be obtained according to Algorithm 2, described below), obtained through a tree structure, is used as the cost of the evaluation and the selection procedure of generating the predicted value is the value of the best estimate of the encoding target image coding. The selected procedure of generating predicted values and some other procedures generate predicted values are saved in the parent population. A new procedure of generating predicted pixel value is automatically generated by repeating the above processes.

Accordingly, the automatic formation procedure of generating predicted values of the pixel that meets the present invention realizes high-precision prediction pixel value by the automatic generation procedure of generating a predicted pixel value based on the way the development tree structure, for example genetic programming. Since the total information content to represent the tree structure and the amount of code estimated predicted value of the pixel obtained through a tree structure, are used as cost estimates, it is possible to automatically generate a procedure to generate a predicted pixel value for the realization of highly efficient coding of images, at the same time preventing the growth of the tree.

In addition, since the method of development of the tree structure is performed under the condition that the current generation function projected values may be a leaf node in the tree, you can achieve the same level of performance prediction, as in the traditional method.

For additional guarantees of obtaining such an effect, when generating a parent population, the parent population can be generated so that the current function of generating predicted values was included in the parent population.

In addition, method development tree structure can be provided that the function that prints the coordinates of the pixel in the image can be a leaf node of the tree. Accordingly, the local switch for the procedure of generating predicted pixel values can be done is manage by using x and y coordinates in accordance with the internal structure of the relevant image.

<2> structure of the device image encoding device and the image decoding according to the present invention (the first type)

When implementing transfer functions automatically generated procedures generate a predicted pixel value decoding side device image coding apparatus and the image decoding in accordance with the present invention have the following structure.

<2-1> Structure of the encoder of the image corresponding to the present invention

When implementing a transfer function generation procedure, the predicted pixel value decoding side, the device coding of images, in accordance with the present invention, has a structure which includes:

(1) the first device, which forms the procedure of generating the predicted values with the value of the best estimate, for the encoding target image coding, through the operations of the automatic formation procedure of generating predicted values of the pixel in accordance with the present invention;

(2) a second device that encodes the procedure of generating the predicted values generated by the first device (using, for example, Algorithm 3 described below);

(3) t the e device which generates the prediction value of each pixel included in the target image coding based on the procedures for generating the predicted values generated by the first device (i.e. using, for example, Algorithm 2, described below); and

(4) a fourth device that encodes the signal residue prediction, calculated using the predicted pixel value generated by the third device.

The method of encoding images that meet present invention, implemented through the operations of the above-described processing devices, can also be implemented through computer program. A computer program is stored on an appropriate computer-readable storage medium or through a network, and the invention is installed and executed on the control device, such as CPU, thanks to which the invention is implemented.

In accordance with the above-described structure of the device for encoding image that meets the present invention realizes high-precision prediction pixel value on the basis of the formation procedure of generating predicted pixel value by the automatic formation that meets the present invention, and encodes the image used is of the generation procedure, the predicted pixel value for the realization of highly efficient coding of images. This enables to realize the high-efficiency coding of images.

<2-2> the Structure of the image decoding conforming to the present invention

For decoding coded data generated by the encoding device image that meets the present invention described in the above paragraph <2-1>, the device decoding images, in accordance with the present invention, has a structure which includes:

(1) the first device that decodes encoded data for the procedure of generating the predicted values generated by the operations of the automatic formation procedure of generating predicted values of the pixel in accordance with the present invention (using, for example, Algorithm 4 described below), where the coded data is generated on the encoding side;

(2) a second device that generates a predicted value of each pixel included in the target image decoding, based on the procedures for generating the predicted values are decoded from the first device (i.e. using, for example, Algorithm 2, described below); and

(3) a third device which decodes encoded data signal residue prediction, computed by the CSO using the predicted pixel value, generated on the basis of the generation procedure, the predicted values are decoded from the first device, where the coded data is generated on the encoding side; and

(4) a fourth device that reproduces the target image decoding based on the predicted pixel value generated by the second device, and signal the rest of the forecast, the decoded third device.

The method of decoding images that meet present invention, implemented through the operations of the above-described processing devices, can also be implemented through computer program. A computer program is stored on an appropriate computer-readable storage medium or through a network, and the invention is installed and executed on the control device, such as CPU, thanks to which the invention is implemented.

In accordance with the above-described structure, the decoding device images that meet present invention, implements the decoding coded data generated by the encoding device image that meets the present invention described in the above paragraph <2-1>.

<3> the Structure of the device image encoding device and the image decoding according to the present invention (the second type)

For each image, encoded at the coding side, the corresponding decoded image is also generated at the decoding side, so the same decoded image can co-exist on the coding and decoding sides. Thus, the transmission procedure of generating predicted pixel value from the encoding side to the decoding side, which is necessary for the implementation of the present invention, can be eliminated.

To implement this for the abolition of the present invention, the device image coding apparatus and the image decoding, in accordance with the present invention have the following structure.

<3-1> Structure of the encoder of the image corresponding to the present invention

When implementing the function of the absence of transfer, the procedure of generating predicted pixel value decoding side device for encoding images, in accordance with the present invention, has a structure which includes:

(1) the first device, which encodes a partial target image coding with a predefined size, using the existing procedures of generating predicted values of the pixel are formed not on the basis of the mode of development of the tree structure is URS;

(2) a second device that provides a procedure for generating the predicted values with the value of the best estimate, for encoding the decoded image obtained during the partial coding target image, encoding the first device, through the operations of the automatic formation procedure of generating predicted values of the pixel in accordance with the present invention, which evaluates the content to represent a tree structure is equal to zero;

(3) the third device that generates a predicted value of each pixel included in the remaining partial target image coding, not encoded by the first device, based on the procedures for generating the predicted values generated by the second device (i.e. using, for example, Algorithm 2, described below); and

(4) a fourth device that encodes the signal residue prediction, calculated using the predicted pixel value generated by the third device.

The method of encoding images that meet present invention, implemented through the operations of the above-described processing devices, can also be implemented through computer program. Computer software is mA is stored on an appropriate computer-readable storage medium or through a network, and the invention is installed and executed on the control device, such as CPU, thanks to which the invention is implemented.

In accordance with the above-described structure of the device for encoding image that meets the present invention realizes high-precision prediction pixel value on the basis of the formation procedure of generating predicted pixel value by the automatic formation that meets the present invention, and performs coding of images using the procedure of generating predicted pixel values for the realization of highly efficient coding of images. This enables to realize the high-efficiency coding of images.

In addition, the above-described device for encoding image that meets the present invention, encodes a partial target image coding with a predefined size, using the existing procedures generate a predicted pixel value generated regardless of the method of development of the tree structure, thereby generating a decoded image for the relevant partial target image coding, where the decoded image can co-exist on the coding and decoding sides of the X. The decoded image is used for the formation procedure of generating predicted pixel values, which may also be formed on the decoding side and has the value of the best estimate. This allows you to do away with the transfer procedure of generating predicted pixel value decoding side.

<3-2> the Structure of the image decoding conforming to the present invention

For decoding coded data generated by the encoding device image that meets the present invention described in the above paragraph <3-1>, the device decoding images, in accordance with the present invention, has a structure which includes:

(1) the first device that decodes encoded data for the partial target image decoding, which has a predefined size, and encoded using an existing procedure of generating predicted values of the pixel are formed not on the basis of the mode of development of the tree structure, where the coded data is generated on the encoding side;

(2) a second device that provides a procedure for generating the predicted values with the value of the best estimate, for encoding partial target image decterov the Oia, received by the first device, through the operations of the automatic formation procedure of generating predicted values of the pixel in accordance with the present invention, which evaluates the content to represent a tree structure is equal to zero;

(3) the third device that generates a predicted value of each pixel included in the remaining partial target image decoding, not decoded by the first device, based on the procedures for generating the predicted values generated by the second device (i.e. using, for example, Algorithm 2, described below);

(4) a fourth device that decodes encoded data signal residue prediction, calculated using the predicted pixel value generated by the generation procedure, the predicted values are decoded from the second device, where the coded data is generated on the encoding side; and

(5) the fifth device that reproduces the remaining partial target image decoding, not decoded by the first device, based on the predicted pixel value generated by the third device, and signal the rest of the forecast, fourth decoded at what trojstva.

The method of decoding images that meet present invention, implemented through the operations of the above-described processing devices, can also be implemented through computer program. A computer program is stored on an appropriate computer-readable storage medium or through a network, and the implementation of the invention, the program is installed and executed on the control device, such as CPU, thanks to which the invention is implemented.

In accordance with the above-described structure, the decoding device images that meet present invention, implements the decoding coded data generated by the encoding device image that meets the present invention described in the above paragraph <3-1>.

In addition, the above-described device, the image decoding conforming to the present invention, decodes the coded data of the partial target image decoding, which has a predefined size, and encoded using an existing procedure of generating predicted pixel value generated regardless of the method of development of the tree structure, thereby generating a decoded image for the relevant partial target image decoding, where decterov the TES, the image may co-exist on the coding and decoding sides. The decoded image is used for the formation procedure of generating predicted pixel values, which may also be formed on the coding side and has the value of the best estimate. This allows you to do away with the transfer procedure of generating predicted pixel value from the coding side.

The result of inventions

As described above, in accordance with the present invention, (i) the procedure for the prediction pixel value is automatically changed using the computer while evaluating content for the procedure of generating predicted pixel value, or (ii) the calculation of the development is also carried out at the decoder side using pixels previously encoded existing method. Thus, you can use the predictor, which can reduce the content of the residue and, thus, to encode the image with the least amount of code.

In addition, since the candidates in the leaf nodes (the present invention) also include the function of the predictor on the basis of traditional methods, the efficiency of forecasting, greater than or equal to (at the bottom), achievable by traditional methods. In addition, candidates may include the coordinates of each pixel to be the kodirovaniyu, that allows you to switch the forecasting procedure in accordance with the internal structure of the relevant image.

Additionally, since the input image can, of course, not to be fixed, it is assumed that evolutionary way of image processing (see Reference document 4 described below)proposed by Nagao and others, in General, be applied to different input images. However, it is virtually impossible to ensure that relevant method can preferably be applied to each unknown input image. In contrast, expectations of the present invention are only focused on efficient encoding the current input image, and to consider such an unknown input is not necessary. Thus, the present invention has a high level of practicality.

In addition, since the total content of the balance and content of the tree is the parameter to be minimize in the present invention, there is no need to use the "training information", which should be prepared by people in General applications of image processing.

Brief description of drawings

Fig. 1 is a diagram showing the view through a tree structure for the averaged prediction.

Fig. 2 - the Hema, explaining the crossover operation.

Fig. 3 is a diagram illustrating the operation of mutation.

Fig. 4 is a diagram illustrating the operation of inversion.

Fig. 5 is a diagram showing the structure of the device forming a developed predictor as an option for implementation of the present invention.

Fig. 6 is a logical block diagram performed by the device for the formation of a developed predictor according to a variant implementation.

Fig. 7 is a diagram illustrating the structure of the encoding device and image decoding device images as an option for implementation of the present invention.

Fig. 8 is a logical flowchart performed by a device for encoding images according to a variant implementation.

Fig. 9 is a logical block diagram executed by the device, the image decoding according to a variant implementation.

Fig. 10 is a diagram illustrating the structure of the encoding device and image decoding device of the image as another variant implementation of the present invention.

Fig. 11 is a logical flowchart performed by a device for encoding images according to a variant implementation.

Fig. 12 is a logical block diagram executed by the device, the image decoding according to a variant implementation.

Fig. 13 is a diagram illustrating the experiment, carried out aemy to test the effectiveness of the present invention.

Fig. 14 is a diagram illustrating a previously decoded pixels that surround the target pixel.

Preferred embodiments of the inventions

The present invention provides for the use of genetic programming (GP) to implement automatic computer forming the forecasting process, which is suitably applied to the input video or static image (hereinafter referred to as simply "image"), and an additional reduction of the relevant amount of code.

Below is the basic idea of the present invention.

<1> View the forecasting process through a tree structure

For example, the average prediction expressed by the above Formula (1)can be represented using a tree structure shown in Fig. 1. For convenience, as a description format that is equivalent to such a presentation of a tree structure, you can use "symbolic expression".

In genetic programming, explained below, a character expression is usually used to represent the tree structure.

For example, in symbolic terms, the above max (x, y) is defined as (max x y), and the above prediction MED is described as follows:

(T (sub (Inw) (max (Iw) (In))) (min (Iw) (In)) (T (sub (min (Iw) (In)) (nw)) (max (Iw) (In)) (add (Iw) (sub (In) (Inw))))),

where each line has a specific value.

Above the T function has three arguments, and the following conditional branching is as follows:

(T A B C)=Bif A≥0
=Cif A <0Formula (5)

where T is the first letter of the word "Trinity".

As described above, any algorithm can be represented in the form of a "tree", and thus, the algorithm for predicting the pixel value can similarly be represented as a tree.

Instead of the above "T" relevant function can use addition, subtraction, multiplication, division, trigonometric function, quadratic function, square root, exponent, logarithm, absolute value, minimum value, maximum value, and so on.

Since this function uses the arguments, it is in a different position than the relevant sheet of wood, and, thus, often referred to as "non-leaf node". The function can be prepared in advance, or can be set dynamically (see Reference document 1).

Reference document 1: J. Koza, "Genetic Programming II, Automatic Discvery of Reusable Programs", The MIT Press, pp. 41, 1998.

In addition, a numerical value, for example, 0,148 or value of the peripheral pixel, for example, Iw, In, Ine or Inw (see Fig. 14) can play the role of the destination node, which itself has value and is assigned to a leaf of a tree.

<2> Characteristics of the leaf nodes in the present invention

In the present invention candidates in the leaf nodes include a function that outputs the predicted value using an existing encoding method.

Since any function needs arguments, this function is not initially assigned to any leaf node. However, the function that outputs the predicted value using an existing encoding method is a function based on an existing encoding method, and thus, the argument types for the function is defined. Thus, the function can also be assigned to the end node.

By analogy with the above value of the peripheral pixel, the predicted value of the output function that outputs the predicted value using an existing encoding method", is determined individually for each target pixel subject to encoding.

The predicted value of the output of the relevant function, may be a predicted value based on Nakanishi the squares, the value obtained planar prediction, the predicted value for CALIC (see Reference document 2 below), the predicted value for JPEG-LS, and so on.

Reference document 2: X. Wu and N. Memon: "Context-Based, Adaptive, Lossless Image Coding", IEEE Transactions on Communications, Vol.45, No.4, pp. 437-444, Apr. 1997.

As described above, when the candidates in the leaf nodes include a function that outputs the predicted value using an existing encoding method, it is possible to achieve the same level of performance prediction, as in the traditional way, essentially without any excessive load.

Thus, in the present invention, as explained below, the procedure of forecasting (i.e. tree structure) for predicting pixel values developed (or evolved) using genetic programming to automatically generate the predictor (i.e. forecasting procedures), which has improved the efficiency of prediction, where the candidates in the leaf nodes include a function that outputs the predicted value using an existing encoding method. Accordingly, the traditional predictor can also be the target for the relevant development.

Thus, if the traditional predictor can provide a higher level of efficiency for predicting the Oia, than the other, automatically generated, predictor, this traditional predictor, in the end, is automatically generated by genetic programming, to achieve the same efficiency prediction, as in the traditional way, essentially without any excessive load.

If the combination of predictor developed (or evolved) using genetic programming, and traditional predictor can provide a more effective forecasting, it is used for encoding.

Additionally, in the present invention candidates in the leaf nodes also include a function that displays the coordinates of the target node, subject to encoding.

Coordinates output from this function can be normalized value, for example "x=-1" for each pixel of the left end; "x=1" for each pixel of the right end, "y=-1" for each pixel of the upper end; "y=1" for each pixel of the lower end of the image, or the actual coordinate values.

Function that displays the coordinates of the target node can randomly display the coordinates in the image plane, without using arguments. Thus, the function can also be assigned to the end node.

As described above, when the candidates of the leaf nodes also include f is ncciu, which displays the coordinates in the image plane) of the target node, subject to encoding, possibly local switching for processing using x and y coordinates, in accordance with the internal structure of the relevant image.

For example, you can build a predictor that performs switching processing in accordance with the value of y so that the top 5/6 part of the image is applied to the predictor, using the procedure of forecasting, and the remaining lower 1/6 part is applied to the predictor that uses a different procedure for forecasting.

<3> estimates for forecasting procedures, the content (amount of information) of the tree and the method of calculating the predicted value

<3-1> estimates for forecasting procedures

With the development (or evolution) of the forecasting process, as explained below, requires evaluation scale.

In the present invention, the sum (X+Y) the following values are used as estimates (referred to as the "degree of agreement" in genetic programming) for each individual, which is the procedure of forecasting:

(i) information content of X (amount of information) to represent the tree structure; and

(ii) the content Y of the remainder of the forecast, the scientists by the actual prediction value of a pixel using forecasting procedures based on the above tree structure.

In the present invention, the estimated value for the individual (the tree is referred to as "individual" in genetic programming) is determined not only based on the content of the Y residue prediction, but also with regard to content X of the tree. One reason is the need to pass procedure of prediction decoding side.

Another reason is that when determining the evaluation value with regard to content X of the tree can prevent the problem of "inflating" (expansion tree) in genetic programming.

<3-2> content X to represent the tree structure

Content X to represent the tree structure is the total content of all tree nodes.

Content to represent the tree structure can be calculated using the following recursive function. It is assumed that a numerical value associated with each node in the alignment of the tree, is expressed, for example, 10-bit integer fixed point.

Algorithm 1

function tree_info(t)
begin
if t is a numeric value then
return FUNCINFO + 10
else begin
s: = FUNCINFO // part of
foreach (all lower nodes c United t) begin
s: = s+tree_info(c)
end
return s
end
end

This assumes that the individual functions are assigned sequence numbers from 0 to N-1.

FUNCINFO has the following value, which indicates the amount of the ode, generated when the function is subjected to encoding with fixed length:

FUNCINFO=log2(N+1) Formula (6),

where (N+1) is also used to account for the numerical values (for example, 2 or 1/4) in addition to the functions.

Although the above discussed coding fixed length, it is possible to perform encoding with variable-length or arithmetic coding, taking into account the frequency of each function.

Then, taking into account the fact that "root" is the highest node in the target procedure predict (tree), the content X of the tree can be calculated as follows:

X=tree_info (root) Formula (7),

<3-3> the Method of calculating the predicted value

The method of calculating the predicted value based on the forecasting process, presents the relevant tree can use a recursive function, as shown below.

Algorithm 2

function tree_eval(t)

begin

if t is a numeric value then

// immediate value

return a numerical value

else if t has no argument then // for example, In

return the value of the function t

else if t has one argument then // for example, sqrt(A)

return function t (tree_eval (first bottom node t))

else if t has two arguments then // for example, add (A, B)

return function t (tree_eval (first bottom node t), tree_eval (second from the bottom node t))

else if t has three arguments then // i.e. the measures the ternary operator T

return function t (tree_eval (first bottom node t), tree_eval (second from the bottom node of t), tree_eval (third from the bottom node t))

end

Although the above algorithm provides that the number of arguments is limited to three or less, similar processes can be carried out even if the upper limit on the number of arguments is equal to 4, 5, ...

Then, taking into account the fact that "root" is the highest node in the target procedure predict (tree), the prediction value p' for the current target pixel can be calculated as follows:

x'=tree_eval (root), Formula (8),

The content Y of the remainder of the forecast can be calculated by the following formula.

[Formula 1]

In the above formula, hdspecifies the number of occurrences (for histograms) prediction error d (= x - x') in the full image, and W and H, respectively, indicates the number of pixels in the horizontal and vertical directions.

Similarly, when performing in CALIC content can be reduced by using a method called "isolation of context", "feedback error" or "transfer error".

<4> the Encoding and decoding procedures of forecasting

<4-1> procedure Coding prediction

Procedure coding prediction also moreviolent using the following recursive procedure, a similar assessment content.

Algorithm 3

procedure tree_encode(t)

begin

if t is a numeric value then begin

to encode the N (numeric value) using bits FUNCINFO

to encode a numerical value (fixed-point) using a 10-bit form.

end else begin

to encode the function number (0,.. N-1) t using bits FUNCINFO

foreach (all lower nodes c United t) begin

tree_encode(c)

end
end
end

After taking into account the fact that "root" is the highest node in the target procedure predict (tree)can be "tree_encode (root), effecting, thus, the relevant coding tree, where the lower limit of the required amount of code coincides with the "tree_info (root)".

<4-2> Decoding procedure of forecasting

The decoding procedure of forecasting, encoded according to Algorithm 3, can also be performed using a similar recursive procedure described below.

Algorithm 4

function tree_decode()

begin

to generate an empty tree T

to decode the bits FUNCINFO to decode functions n

if n = N then begin // numerical value

to decode 10 bits to decode 10-pitogo x values are fixed-point

T:= x

end else begin

to calculate the function F corresponding to the function number n

T:= F

for i = 1 to "number of arguments required for F begin

the i-th bottom node T:= tree_decode()

end
end
return T
end

The above number of arguments required for F, is the number of (known coding and decoding sides) values used for output values of the relevant functions. If F=add relevant number is equal to 2, and if F=T, the relevant number is equal to 3.

Here F is the lower nodes corresponding to the relevant number, as their own arguments.

Then, when you run tree_decode() tree is decoded by the relevant bit stream and then returned.

<5> Automatic development of procedures for forecasting using genetic ol the program

In the present invention, the predictor is developed through the following well-known procedures (including the choice of copies, the generation of offspring and selection of survivors) for genetic programming.

In genetic programming, each tree is called the "individual". This is based on the following explanation.

1. First of all, pre-generated population using random numbers or the existing prediction algorithm (e.g., above the plane or MED forecasting).

2. From the population is selected many parents (parent population) (select copies).

3. From the parent population is generated (generation descendants) and evaluated (evaluation scale explained above) the number of children individuals.

4. Based on the results of the assessment are selected survivors of the number of daughter individuals (selection survivors).

According to the above procedure, each offspring is generated by implementing the following process between individuals selected as parents:

(i) the crossover shown in Fig. 2, for the random selection of crossover points in the parents 1 and 2, and the implementation of crossover between parent trees in accordance with the points of crossing-over;

(ii) mutation, shown in Fig. 3, for the random selection of a point mutation, and replacement of the ro is itelescope tree mutated tree in accordance with point mutations; or

(iii) the inversion shown in Fig. 4, for the exchange between the trees-brothers.

The choice of copies and select the survivors are together called "the model of alternation of generations", for which you can apply a well-known method MGG (minimum gap between generations), proposed in the following Reference document 3.

Reference document 3: Hiroshi Sato, Isao Ono, and Shigenobu Kobayashi, "A New Generation Alternation Model for Genetic Algorithms and Its Assessment", Journal of Japanese Society for Artificial Intelligence, Vol. 12, No. 5, pp. 734-744, 1996.

In the following Reference document 4 representative revealed how the development of operating procedures using genetic programming, in a way that relates to the procedure for image processing.

Reference document 4: Wataru Fujishima and Tomoharu Nagao, "PT-ACTIT; Tunable Parameter-Automatic Construction of Tree-structural Image Transformation", Journal of Institute of Image Information and Television Engineers, Vol. 59, No. 11, pp. 1687-1693, 2005.

However, it was not proposed method development procedure coding "images" (provided by the present invention). The above methods disclosed in non-Patent document 4 or 5 are just an optimization parameter encoding.

Below, the present invention will be explained in detail using embodiments.

In Fig. 5 shows the structure of the device 1 forming a developed predictor as an option for implementation of altoadige of the invention.

The device 1 forming a developed predictor in the present embodiment, implements the automatic generation of the predictor, which uses genetic programming (in which each tree is referred to as an individual) to generate predicted pixel values. Next, the predictor formed in the present embodiment, referred to as the developed predictor.

To implement automatic generation, as shown in Fig. 5, the device includes a block 10 generating a parent population, unit 11 to store the parent population, the block 12 selection and duplication of the parent of the individual, unit 13 generating subsidiary of individuals, the storage unit 14 mutational information, the calculation unit 15 estimates the block 16 definition of tenacious individuals, block 17 definition of convergence, and the block 18 definition of a developed predictor.

Unit 10 generating a parent population to generate the parent population, randomly generating individuals for predictor as the initial option for a developed predictor, and saves the parent population in unit 11 to store the parent population. In this process, the existing function of generating predicted values (as an individual) is contained in the generated and stored parent population.

Unit 10 generation bear liskay population also requests the block 15 estimated values to calculate an estimated value for each individual, stored in the storage block 11 parent population, and receives an estimated value returned from the block 15 estimated values in response to a relevant query. Unit 10 generating a parent population saves each estimated value in unit 11 to store the parent population, in connection with the appropriate individual (saved).

Unit 12 selection and duplication of the parent of the individual selects and duplicates the totality of individuals stored in the storage block 11 parent population, thereby generating a set of parent individuals.

Unit 12 selection and duplication of the parent of the individual also removes individuals as a source of options for the generated parent individuals from the unit 11 to store the parent population.

On the basis of genetic programming unit 13 generating subsidiary of individuals generates a subsidiary of individuals, selecting parent individuals generated by the block 12 selection and duplication of the parent of the individual, for crossover, shown in Fig. 2; mutations shown in Fig. 3, using mutational information stored in the storage unit 14 mutational information; inversion shown in Fig. 4; and so on.

Unit 13 generating subsidiary of individuals also calculates an evaluation value for each generated dozer is his individual requesting unit 15 estimated values to calculate the estimated value and taking the estimated value returned from the block 15 estimated values in response to the relevant request.

Unit 14 storage mutation information mutation retains information (i.e. mutated tree), used when the unit 13 generating subsidiary of individuals exposed mutations parent individual, and mutational information includes a function (as an individual), which displays the relevant coordinates in the image and the existing generating function of the predicted values (as an individual).

When the calculation unit 15 estimates accepts the request to calculate the estimated value for a given individual, it calculates the total aggregate content (above X: the content of the individual), are required to represent the corresponding tree structure, and content (above Y: the content of the remainder of the forecast) balance forecasting full image, for which the prediction pixel value was actually carried out using the procedure of forecasting on the basis of the relevant tree structure. Block 15 estimated values returns the computed total full and the formational content as estimated values for the individual, on the unit that issued the query, calculate the estimated value.

On the basis of estimated values (extracted from the unit 11 to store the parent population) for each parent of the individual, generated by the block 12 selection and duplication of the parent of the individual, and the evaluation values assigned to each child individual generated by the unit 13 generating subsidiary of individuals, block 16 definition of tenacious individuals selects the individual with the best estimated value, and stores the selected individual and one or more other individuals in the unit 11 to store the parent population, together with the corresponding estimated values.

On the basis of the evaluation values outputted from the calculation unit 15 estimates and area, block 17 definition of convergence determines whether the convergence condition, which indicates the completion of the formation of a developed predictor. If it is determined that the condition is met, the block 17 definition of convergence instructs the block 18 definition of a developed predictor to determine a developed predictor.

Taking the definition query of the developed predictor from block 17 definition of convergence, block 18 definition of a developed predictor indicates the individual having the best evaluation value among the individuals stored in the storage block 11 parent population, and to define the em and displays the specified individual as the developed predictor.

In Fig. 6 shows a logical block diagram performed by the device 1 forming a developed predictor of having above-described structure.

In accordance with a logical block diagram will be explained in detail the operation performed by the device 1 forming a developed predictor.

According to logic flow diagram depicted in Fig. 6, having made the request to generate a developed predictor for the image as the target of the encoding device 1 forming a developed predictor of first generates the parent population (i.e. many individuals as the source variation for the relevant development), which includes the individual, which displays the predicted value using an existing encoding method (see step S101).

In the next step S102 for each individual in the parent population, the total information content of X to represent the corresponding tree and the content Y for the remainder of the forecast the entire image, for which the prediction pixel value was actually carried out using the procedure of forecasting on the basis of the relevant tree structure, is calculated to calculate the estimated value.

Content X to represent the corresponding tree is calculated using wireapi the data algorithm 1.

The predicted value obtained by the procedure of forecasting on the basis of the tree structure is computed using the above algorithm 2.

In the next step S103, each individual in the parent population is stored in the storage block 11 parent population, together with the estimated value assigned to the individual.

In the next step S104 N parent individuals are selected among the individuals stored in the storage block 11 parent population, and assigned estimated values are also retrieved.

In the next step S105 selected N individuals are duplicated and also removed from the unit 11 to store the parent population.

In the next step S106 M child individuals are generated from the N duplicated parent individuals by implementing, for example, crossover, shown in Fig. 2, the mutations shown in Fig. 3, or inversion, is shown in Fig. 4, by means of genetic programming, which can be used mutational information stored in the storage unit 14 mutational information.

In the above process, the candidates individuals (trees), added by mutations include a function for generating the predicted values using the traditional method, and the function that prints the x and y coordinates of the pixel to be what its encoding.

In the next step S107 for each of the generated M child individuals total information content of X to represent the corresponding tree structure and information content of Y for the remainder of prediction actual prediction pixel value using the procedure of forecasting on the basis of the relevant tree structure is computed to calculate the estimated value.

In the next step S108 among the goals of choice, consisting of M generated affiliated individuals and duplicate N parent individuals, selects the individual with the best estimated value, and the other N-1 individuals are randomly chosen, as tenacious individuals.

In the next step S109 selected tenacious individuals stored in the storage block 11 parent population with the relevant assigned estimated values.

In the next step S110, it is determined whether a predetermined convergence condition. If it is determined that the convergence condition is not yet completed, it has been decided that the development at the moment is not enough, and the operation returns to step S104.

Used the convergence condition may be that at which the rate of reduction of the estimated value Z (X+Y) is less than a fixed value (e.g., 0.1 percent), and the number of iterations to calculate the estimated value exceeds a fixed value (for example, 10,000).

If the above step S110 decided that a predetermined convergence condition is fulfilled, the operation proceeds to step S111, where the individual having the highest evaluation value is selected and displayed as finally evolved individual (i.e. a developed predictor), among individuals of the parent population that is stored in the storage block 11 parent population. This operation completes.

As described above, the device 1 forming a developed predictor according to the present variant implementation can automatically form a developed predictor, which realizes high-precision prediction pixel value, using genetic programming.

To implement such an automatic shaping device 1 forming a developed predictor according to the present variant implementation uses the estimated value that is the total content X to represent a tree structure, and the content Y for the remainder of the forecast the entire image, for which the prediction pixel value was actually carried out using the procedure of forecasting on the basis of the relevant tree structure. Thus, it is possible to automatically generate the predictor, which Khujand which performs prediction pixel value for the realization of highly efficient coding of images.

For the above operation, the candidates added individuals include a function that generates a predicted value, using the traditional method. Thus, it is possible to achieve the same level of performance prediction, which provides the traditional method.

In addition, when such an individual as a function that generates a predicted value using the traditional method, is included in the parent population at generation parent population, the aforementioned "the same level of performance prediction" can be implemented more confident.

In addition, since the candidates to added individuals include the function that prints the x and y coordinates of the pixel subject to encoding is also possible to make a local change in the developed predictor in accordance with the internal structure of the relevant image using the x and y coordinates.

In Fig. 7 shows embodiments of device 100 encoding images and device 200 decoding images that use the device 1 forming a developed predictor according to the present variant implementation.

The device 100 coded image shown in Fig. 7, includes a block 101 of forming a developed predictor, which is no developed predictor, applied to the target image encoding in accordance with the operation performed by the device 1 forming a developed predictor according to the above variant implementation, the block 102 encoding a developed predictor for coding a developed predictor formed by block 101 forming a developed predictor, block 103 encoding images to encode the target image coding using a developed predictor formed by block 101 forming a developed predictor, and the block 107 transmission of the coded data, which transmits to the device 200 of the image decoding (in the present embodiment) of the coded data generated by the block 102 encoding a developed predictor and the block 103 encoding of images.

The above block 103 encoding of images includes generator 104 predicted pixel value, which predicts the pixel value using the developed predictor formed by block 101 forming a developed predictor, block 105 calculation of residue prediction, which computes the remainder of forecasting on the basis of the pixel values predicted by the generator 104 predicted pixel value, and the encoder 106 residue prediction for coding of the residual prediction computed by block 105 in which the calculations residue prediction.

The device 200 of the image decoding shown in Fig. 7, includes unit 201 receiving encoded data for receiving encoded data transmitted from the device 100 encoding images according to the present variant implementation, the decoding unit 202 of the developed predictor for decoding a developed predictor generated by the device 100 encoding image by decoding encoded data of a developed predictor adopted by the unit 201 receiving encoded data, and unit 203 of the image decoding for decoding the image encoded by the device 100 coding of images, on the basis of the developed predictor, decoded by the decoding unit 202 of the developed predictor, and the coded data received by the unit 201 receiving encoded data.

For decoding the image encoded by the device 100 image coding, unit 203 of the image decoding includes generator 204 predicted pixel value for predicting the predicted values using the developed predictor, decoded by the decoding unit 202 of the developed predictor, the decoder 205 residue prediction for decoding encoded data of the remainder of the forecast adopted by the unit 201 receiving encoded data, and the player 206 image is s to display the image, encoded device 100 encoding of images based on the pixel value, the predicted generator 204 predicted pixel value, and the remainder of the forecast, decoded by the decoder 205 residue prediction.

In Fig. 8 shows a logical block diagram performed by the device 100 encoding of images is shown in Fig. 7, and in Fig. 9 shows a logical block diagram performed by the device 200 of the image decoding shown in Fig. 7.

In accordance with a logical block diagram will be explained the operations of the device 100 image encoding device 200 of the image decoding having the structure shown in Fig. 7.

According to logic flow diagram depicted in Fig. 8, when the device 100 coding of images having the structure shown in Fig. 7, receives the request for the encoding target image coding, it first generates a developed predictor applied to the target image coding, on the basis of the operations performed by the above-described device 1 forming a developed predictor (see step S201). In the next step S202 formed developed a predictor is encoded using the above algorithm 3.

Then, to encode the target image coding predicted value of the pixel is La (above p') is generated using a set of developed predictor (see step S203), and then the remainder of the forecast (above p-p') is calculated based on the generated predicted value of the pixel (see step S204).

In the next step S205 the remainder computed prediction is encoded, and in the next step S206, it is determined, whether completed coding for all pixels contained in the target image coding. If it is determined that the encoding of all pixels has not been completed, the operation returns to step S203. If it is determined that the encoding of all pixels is completed, the current operation is terminated.

According to logic flow diagram depicted in Fig. 9, when the device 200 decodes images having the structure shown in Fig. 7, receives the coded data generated by the device 100 of the image encoding device 200 of the image decoding first decodes the coded data of the developed predictor based on the above algorithm 4, to decode the developed predictor generated by the device 100 image encoding (see spider S301 demonstration stage).

Upon further decoding target image decoding phase W302, the predicted pixel value (above p') is generated using the decoded developed predictor, and in the next step S303 coded data balance forecast the simulation are decoded to obtain the decoded residue prediction (i.e. the above p-p'). In the next step S304, the pixel value is generated and displayed based on the previously generated predicted pixel value and the decoded residue prediction.

In the next step S305, it is determined, whether completed relevant decoding for all pixels included in the target image decoding. If it is determined that the decoding of all pixels has not been completed, the operation returns to step W302. If it is determined that the decoding of all pixels is completed, the current operation is terminated.

As described above, the device 100 coding of images having the structure shown in Fig. 7, forms a developed predictor, encodes the image using the developed predictor and also encodes a developed predictor. The device 200 decodes images having the structure shown in Fig. 7, receives the developed predictor generated by the device 100 encoding image by decoding encoded data of a developed predictor and decodes the relevant image using the developed predictor.

Also, as described above, the device 1 forming a developed predictor automatically generates a developed predictor, which realizes high-precision prediction of pixel values.

Thus, in accordance with the laws the AI with the device 100 of the image encoding device 200 of the image decoding, which encode and decode the image using the developed predictor generated by the device 1 forming a developed predictor, it is possible to achieve high coding efficiency.

In Fig. 10 shows other embodiments of device 100 encoding images and device 200 decoding images that use the above described device 1 forming a developed predictor.

The device 100 encoding image and the device 200 of the image decoding shown in Fig. 7, require encoding and transfer of the developed predictor. However, the device 100' encoding image and the device 200' decoding of images shown in Fig. 10, do not perform such encoding and transfer of the developed predictor, and a developed predictor formed the coding side, can be formed decoder side by using the previously transmitted pixel, allowing encoding and decoding side can encode and decode the image using the same advanced predictor.

For realizing the aforesaid functions of the device 100' encoding is shown in Fig. 10, has a first encoding unit 110 of the image to encode a partial image that is the target image codero the project and has a predefined size, using an existing predictor; block 111 forming a developed predictor for the formation of a developed predictor applied to the decoded image of an encoded partial image where the decoded image obtained by decoding performed by the first encoding unit 110 of the image, for the implementation of the relevant coding; the second coding block 112 to encode the remaining partial images of the target image coding using a developed predictor formed by block 111 forming a developed predictor; and block 116 transmission of the coded data, which transmits the coded data generated by the first coding block 110 and the second coding block 112, the device 200' decoding of images.

To encode the partial images that are not encoded by the first coding block 110, the above-mentioned second coding block 112 includes a generator 113 predicted pixel value, which predicts the pixel value using the developed predictor formed by block 111 forming a developed predictor, block 114 calculation of residue prediction, which computes the remainder of forecasting on the basis of the pixel values predicted by the generator 113 predicted value of the pixel, and the encoder 115 residue prediction for coding of the residual prediction calculated by the calculation block 114 residue prediction.

The device 200' decoding of images shown in Fig. 10, has unit 210 of receiving encoded data for receiving encoded data transmitted from the device 100' coding of images; the first unit 211 of the image decoding to decode encoded data generated by the first encoding unit 110 of the image contained in the coded data received by the unit 210 of receiving encoded data; block 212 forming a developed predictor for the formation of a developed predictor applied to the partial image, the decoded first block 211 decoding images; the second block 213 of the image decoding to decode the partial image is not decoded first block of the image decoding 211, on the basis of the developed predictor formed by block 212 of forming complex predictor, and the encoded data generated by the second coding block 112 and is contained in the encoded data received by the unit 210 of receiving encoded data; and a synthesizer 217 image to generate image as a target of decoding by the synthesis image, the decoded first block 211 decoding picture is, and the image decoded by the second unit 213 decoding of images.

To decode the partial image is not decoded first block 211 decoding images, the above-mentioned second block 213 of the image decoding includes generator 214 predicted pixel value for predicting the predicted values using the developed predictor formed by block 212 forming a developed predictor; the decoder 215 residue prediction for decoding encoded data of the remainder of the forecast adopted by the unit 210 of receiving encoded data and the designated partial image which is not decoded first block 211 decoding images; and player 216 images for reproducing a partial image, which is not decoded first block 211 decoding image based on the pixel value, the predicted generator 214 the predicted pixel value, and the remainder of the forecast, decoded by the decoder 215 residue prediction.

In Fig. 11 shows a logical block diagram performed by the device 100' coding of images, shown in Fig. 10, and in Fig. 12 shows a logical block diagram performed by the device 200' decoding images, shown in Fig. 10.

In accordance with the logical whom they block diagram will be explained the operation, performed by the device 100' image encoding device 200' decoding of images.

When the device 100' encoding image receives the request for encoding the image, it starts with encoding the partial image having N pixels as the intended conditions), which belongs to the target image coding and has a predefined size, using an existing predictor (see step S401). In the next step S402, the encoding continues until it is confirmed by the completion of coding for the relevant partial image and, thus, the encoding of the partial image.

For example, such partial image belonging to the target image encoding and which has a predefined size, coded using JPEG-LS.

In the next step S403 developed predictor applied to the decoded image (encoded partial image data)obtained by encoding the above-described step S401, is formed on the basis of the operations performed by the above-described device 1 forming a developed predictor.

As described above, basically, the developed predictor, with a preferred estimate, is formed in accordance with the estimated value specified in the image quality is as total content X to represent the tree structure and the content of Y for the remainder of the forecast the entire image, for which the prediction pixel value was actually carried out using the procedure of forecasting on the basis of the relevant tree structure. However, since in the present embodiment, is not required to transfer the developed predictor, the estimated value is calculated by setting the content X of 0 (the procedure of development does not change), and developed a predictor is formed on the basis of the calculated evaluation values.

Then, to encode the remaining partial images, which also belongs to the target image coding, but not encoded on the above-described step S401, the predicted pixel value (above p') is generated using a set of developed predictor (see step S404), and in the next step S405 balance forecast (above p-p') is calculated based on the generated predicted pixel value.

In the next step S406 the remainder computed prediction is encoded, and the next step S407, it is determined, whether completed coding for all pixels contained in the target image coding. If it is determined that the encoding of all pixels has not been completed, the operation returns to step S404. If it is determined that the encoding of all pixels is completed, the current operas is tion is terminated.

According to logic flow diagram depicted in Fig. 12, when the device 200' decoding takes the image coded data generated by the device 100' image encoding device 200' of the image decoding starts with decoding encoded data of a partial image having N pixels as the intended conditions), encoded by the device 100' coding of images using an existing predictor (see step S501). In the next step S502, the decoding continues until it is confirmed by the completion of decoding for the relevant partial image and, thus, the partial decoding of the image.

For example, a partial image having a predefined size, is decoded using JPEG-LS.

In the next step S503 developed predictor applied to the decoded image obtained by decoding in the above step S501, is formed on the basis of the operations performed by the above-described device 1 forming a developed predictor.

As described above, basically, the developed predictor, with a preferred estimate, is formed in accordance with the evaluation value is defined as the sum of the content X to represent the tree is structure and content Y for the remainder of the forecast the entire image, for which the prediction pixel value was actually carried out using the procedure of forecasting on the basis of the relevant tree structure. However, since in the present embodiment, is not required to transfer the developed predictor, the estimated value is calculated by setting the content of X is equal to 0 (the procedure of development does not change), and developed a predictor is formed on the basis of the calculated evaluation values.

Then, to decode the remaining partial images, which also belongs to the target image decoding, the predicted pixel value (above p') is generated using a set of developed predictor (see step S504), and in the next step S505 balance forecast (above p-p') is decoded by decoding encoded data of the residual prediction.

In the next step S506, the pixel value is generated and displayed based on the previously generated predicted pixel value and the decoded residue prediction.

In the next step S507, it is determined, whether completed relevant decoding for all the pixels included in the remaining partial image of the target image decoding. If it is determined that the decoding of all the pixel is not completed yet, the operation returns to step S504. If it is determined that the decoding of all pixels is completed, the current operation is terminated.

As described above, the device 100' encoding of the image having the structure shown in Fig. 10 encodes the portion of the target image coding using an existing encoding that generates a developed predictor using the decoded image obtained in this encoding, and encodes the remaining partial image formed using the developed predictor.

The device 200' decoding images having the structure shown in Fig. 10, decodes the portion of the target image decoding by decoding the relevant coded data in accordance with the existing method of decoding, generates a developed predictor using the decoded image, and decodes the remaining partial image formed using the developed predictor.

In addition, as described above, the device 1 forming a developed predictor automatically generates a developed predictor, which realizes high-precision prediction of pixel values.

Thus, in accordance with the device 100' image encoding device 200' decoding images that encode and Dec is dirout the image using the developed predictor, generated by the device 1 forming a developed predictor, it is possible to achieve high coding efficiency.

In the experiment conducted by the authors of the present invention for testing the effectiveness of the invention, when the estimated value of X+Y is minimized for images, you can generate the following relatively simple developed predictor.

(add (sub 0.5 (sub (div (Igap) (Ine)) (Igap))) (div (Inw) (Igap))),

where Igap denotes nonlinear prediction value obtained from the periphery, and Ine and Inw are the values of the peripheral pixels, as shown in Fig. 14.

Estimate the value of X+Y had the size 1170235 bits, which is better than the maximum value 1176090 bits obtained using currently available existing predictors.

Relevant advanced predictor performs division between the predicted values and this indicates that the present invention enables the forming of generating predicted values, which cannot be expected, considering the traditional predictors.

Below are the results of an experiment carried out to test the effectiveness of the invention. In the experiment the current function of generating predicted values (as an individual) was not included in the parent population.

In this experiment compare the nutrient trees forecasting was a predictor based on the least squares (LS), which performs linear prediction, the predictor based on the minimum entropy (LE), which also performs linear prediction, the predictor GAP for CALIC, which performs linear prediction (using the four peripheral pixels, similar to the present invention (see Fig. 14)), and the MED predictor for JPEG-LS, which also performs non-linear forecasting (using three peripheral pixels).

When the above-mentioned prediction "LE" five factors in the prediction of LS are used as initial values, and Y is minimized through a multidimensional search based on Powell's method. Forecasting LE provides the highest level of performance among the methods of linear prediction.

In Fig. 13 shows the content (X+Y) for the remainder of each image (with 512×512 pixels, 8-bit gradation and only the brightness data)used in the experiment, where the content also takes into account the service load (50 to LS and LE, 0 for MED, and the bits of X for the method proposed according to the present invention). In Fig. 13 also shows each increase with respect to the content of the proposed method for each image. In addition, in the bottom row of Fig. 13 shows the content X of the tree is, meets the present invention, for each image. Unit residual content is bpp (pits per pixel).

The results of the experiment confirmed that the predictor that is automatically generated in accordance with the present invention, has the highest level of efficiency. The average content X for tree predictor, automatically generated in accordance with the present invention, is 726 bits, which indicates a small complication compared with predictors GAP and MED (which, respectively, have 349,5 bits and 116,0 bits, and set equal to zero in this experiment).

For the image "Lena" calculation of the development was carried out separately without considering the content X of the tree to minimize only the residual information content of y In comparison with the results presented in Fig. 13, the results of separate calculations for the proposed method that meets the present invention, show that Y (the value of Y is not shown in Fig. 14) decreased by 0.06%, but X has increased by about three times (i.e. X=2795 bits), so that X+Y has increased by 0.14%. Although the growth of the tree is the problem, referred to as "bloat", due to the SE, the present invention is, of course, prevents its occurrence is due to the couple X.

For the image "Baboon" was formed tree prediction for different destination processes the bottom 1/6 of the area and the remaining area in the image. The area is 1/6 and the remaining area corresponds to whether it is an area that has only a beard. This tree prediction supports the high-level search on the basis of SOEs.

The effectiveness of the present invention can be verified on the basis of the above experimental results.

Industrial application

As described above, the present invention can be applied to encoding and decoding video or a static image, to implement high-precision prediction pixel value. Thus, using the computer can automatically generate the forecasting procedure, suitable for each input image, and to further reduce the relevant amount of code.

The list of symbols

1 device for the formation of a developed predictor

10 unit generating a parent population

11, the storage unit parent population

12 unit selection and duplication of the parent of the individual

13 block generation subsidiary of individuals

14, the storage unit mutational information

15 unit calculating the estimated value

16 block definition tenacious individuals

17 block determine the expenditure is the cost

18 block definition developed predictor

1. A method of coding images for encoding an image using the predicted pixel value generated by a predetermined procedure of generation of the predicted values, which predicts the value of the target pixel encoding using a previously decoded pixels, and the method includes
the first stage, and when you accept the request for the encoding target image encoding, at the first stage form the procedure of generating the predicted values with the value of the best estimate, for the encoding target image coding by a predetermined method automatic generation procedure of generating predicted values of the pixel
the second phase, which encode the procedure of generating the predicted values generated in the first stage,
the third stage, which generates the prediction value of each pixel included in the target image coding based on the procedures for generating the predicted values generated in the first stage,
the fourth stage, which encode the signal residue prediction for each pixel, and the above-mentioned signal is calculated based on the predicted pixel value generated by the CSOs in the third stage, and
the fifth stage, which send the data encoded on the second and fourth stages,
moreover, the above-mentioned method of automatic generation procedure of generating predicted pixel values contains:
the first stage, which generate the parent population through random generation procedure of generating predicted values, each of which is specified tree structure,
the second stage, which selects the set of procedures for generating the predicted values as parents from the parent population to form one or more of the procedures generating the predicted values as descendants on the basis of a predetermined method development tree structure, which puts the development of the selected procedure of generating predicted values, where the existing generation function projected values may be a leaf node in the tree,
the third stage, which
choose the procedure of generating predicted values having a minimum value of an evaluation of the procedures generating the predicted values as the parents and children, where the total information content to represent the tree structure and the amount of code estimated predicted value of the pixel obtained through a tree structure, using the t as cost estimates, and selected procedure for generating the predicted value is the value of the best estimate of the encoding target image encoding, and
save the selected procedure of generating predicted values and one or more other procedures generate predicted values in the parent population, and
the fourth stage, which manage to repeat the second and third steps until the predetermined condition, and form the process of generation of the predicted values with the value of the best estimate, as a result of repetition as the final procedure of generating predicted values.

2. The method of encoding images according to claim 1, in which
in the second stage of the above-mentioned method of automatic generation procedure of generating predicted values of the pixel generation procedure predicted values as descendants form the basis of a predetermined method development tree structure, which carries out the development, where the function that prints the coordinates of the pixel in the image can be a leaf node of the tree.

3. The method of encoding images according to claim 1, in which
at the first stage of the above-mentioned method of automatic generation procedure of generating predicted pixel values Rodi is Yelsk population generate so to the existing function of generating predicted values was included in the parent population.

4. The way the image decoding to decode encoded image data, encoded predicted pixel value generated by a predetermined procedure of generation of the predicted values, which predicts the value of the target pixel encoding using a previously decoded pixels, and the method includes
the first stage, which receive and decode the encoded data generation procedure, the predicted values generated by a predetermined method automatic generation procedure of generating predicted pixel values, where the coded data is generated on the encoding side,
the second stage, which generates the prediction value of each pixel included in the target image decoding based on the procedures for generating the predicted values decoded in the first stage, and
the third stage, which decode the encoded data signal residue prediction for each pixel, and the signal is calculated using the predicted pixel value generated by the generation procedure to the formation the constituent values, decoded in the first stage, where the coded data is generated on the encoding side, and
the fourth stage, which reproduce a target image decoding based on the predicted pixel value generated in the second stage and the signal residue prediction decoded at the third stage,
moreover, the above-mentioned method of automatic generation procedure of generating predicted pixel values contains:
the first stage, which generate the parent population through random generation procedure of generating predicted values, each of which is specified tree structure,
the second stage, which selects the set of procedures for generating the predicted values as parents from the parent population to form one or more of the procedures generating the predicted values as descendants on the basis of a predetermined method development tree structure, which puts the development of the selected procedure of generating predicted values, where the existing generation function projected values may be a leaf node in the tree,
the third stage, which selects the procedure of generating predicted values having a minimum value of an evaluation of the procedures generating the predicted C is achene as parents and descendants, where the total content to represent the tree structure and the amount of code estimated predicted value of the pixel obtained through a tree structure, used as a cost estimate, and the selected procedure generating the predicted value is the value of the best estimate of the encoding target image encoding, and
save the selected procedure of generating predicted values and one or more other procedures generate predicted values in the parent population, and
the fourth stage, which manage to repeat the second and third steps until the predetermined condition, and form the process of generation of the predicted values with the value of the best estimate, as a result of repetition as the final procedure of generating predicted values.

5. The way the image decoding according to claim 4, in which
in the second stage of the above-mentioned method of automatic generation procedure of generating predicted values of the pixel generation procedure predicted values as descendants form the basis of a predetermined method development tree structure, which carries out the development, where the function that prints the coordinates of the pixel in the image, can be a leaf node of the tree.

6. The way the image decoding according to claim 4, in which
at the first stage of the above-mentioned method of automatic generation procedure of generating predicted pixel values of the parent population to generate such a way that the current generation function of the predicted values was included in the parent population.

7. A method of coding images for encoding an image using the predicted pixel value generated by a predetermined procedure of generation of the predicted values, which predicts the value of the target pixel encoding using a previously decoded pixels, and the method includes
the first stage, and when you accept the request for the encoding target image coding, in the first phase encode partial target image coding with a predefined size, using the existing procedures of generating predicted values of the pixel are formed not on the basis of the mode of development of the tree structure
the second stage, which form the procedure of generating the predicted values with the value of the best estimate, for encoding the decoded image obtained during encoding partial target from the expression of the coding in the first stage, by a predetermined method automatic generation procedure of generating predicted pixel values, where appreciate the content to represent a tree structure is equal to zero,
the third stage, which generates the prediction value of each pixel included in the remaining partial target image coding, not encoded in the first stage, based on the procedures for generating the predicted values generated in the second stage;
the fourth stage, which encode the signal residue prediction for each pixel, and the above-mentioned signal is calculated based on the predicted pixel value generated in the third step; and
the fifth stage, which send the data encoded in the first and fourth stages,
moreover, the above-mentioned method of automatic generation procedure of generating predicted pixel values contains:
the first stage, which generate the parent population through random generation procedure of generating predicted values, each of which is specified tree structure,
the second stage, which selects the set of procedures for generating the predicted values as parents from the parent population to form one or more generators which of the predicted values as descendants on the basis of a predetermined method development tree structure, which puts the development of the selected procedure of generating predicted values, where the existing generation function projected values may be a leaf node in the tree,
the third stage, which
choose the procedure of generating predicted values having a minimum value of an evaluation of the procedures generating the predicted values as the parents and children, where the total information content to represent the tree structure and the amount of code estimated predicted value of the pixel obtained through a tree structure, used as a cost estimate, and the selected procedure generating the predicted value is the value of the best estimate of the encoding target image encoding, and
save the selected procedure of generating predicted values and one or more other procedures generate predicted values in the parent population, and
the fourth stage, which manage to repeat the second and third steps until the predetermined condition, and form the process of generation of the predicted values with the value of the best estimate, as a result of repetition as the final procedure of generating predicted values.

8. The coding method from the of interests according to claim 7, in which
in the second stage of the above-mentioned method of automatic generation procedure of generating predicted values of the pixel generation procedure predicted values as descendants form the basis of a predetermined method development tree structure, which carries out the development, where the function that prints the coordinates of the pixel in the image can be a leaf node of the tree.

9. The method of encoding images according to claim 7, in which
at the first stage of the above-mentioned method of automatic generation procedure of generating predicted pixel values of the parent population to generate such a way that the current generation function of the predicted values was included in the parent population.

10. The way the image decoding to decode encoded image data, encoded predicted pixel value generated by a predetermined procedure of generation of the predicted values, which predicts the value of the target pixel encoding using a previously decoded pixels, and the method includes
the first stage, which receive and decode the coded data for the partial target image decoding, which has a predefined RA the measures and encoded using an existing procedure of generating predicted values of the pixel are formed not on the basis of the mode of development of the tree structure, where the coded data is generated on the encoding side,
the second stage, which form the procedure of generating the predicted values with the value of the best estimate, for encoding partial target image decoding obtained at the first step, by a predetermined method automatic generation procedure of generating predicted pixel values, where appreciate the content to represent a tree structure is equal to zero,
the third stage, which generates the prediction value of each pixel included in the remaining partial target image decoding, not decoded in the first stage, based on the procedures for generating the predicted values generated in the second stage,
the fourth stage, which decode the encoded data signal residue prediction for each pixel, and the signal is calculated using the predicted pixel value generated based on the procedures for generating the predicted values decoded in the second stage, where the coded data is generated to drowsey side, and
the fifth stage, which reproduce the remaining partial target image decoding, not decoded in the first stage, based on the predicted pixel value generated in the third stage, and the signal residue prediction decoded in the fourth stage,
moreover, the above-mentioned method of automatic generation procedure of generating predicted pixel values contains:
the first stage, which generate the parent population through random generation procedure of generating predicted values, each of which is specified tree structure,
the second stage, which selects the set of procedures for generating the predicted values as parents from the parent population to form one or more of the procedures generating the predicted values as descendants on the basis of a predetermined method development tree structure, which puts the development of the selected procedure of generating predicted values, where the existing generation function projected values may be a leaf node in the tree,
the third stage, which
choose the procedure of generating predicted values having a minimum value of an evaluation of the procedures generating the predicted values as parents and flux the Cove, where the total content to represent the tree structure and the amount of code estimated predicted value of the pixel obtained through a tree structure, used as a cost estimate, and the selected procedure generating the predicted value is the value of the best estimate of the encoding target image encoding, and
save the selected procedure of generating predicted values and one or more other procedures generate predicted values in the parent population, and
the fourth stage, which manage to repeat the second and third steps until the predetermined condition, and form the process of generation of the predicted values with the value of the best estimate, as a result of repetition as the final procedure of generating predicted values.

11. The method of decoding images of claim 10, in which
in the second stage of the above-mentioned method of automatic generation procedure of generating predicted values of the pixel generation procedure predicted values as descendants form the basis of a predetermined method development tree structure, which carries out the development, where the function that prints the coordinates of Pixela image, can be a leaf node of the tree.

12. The method of decoding images of claim 10, in which the first stage of the above-mentioned method of automatic generation procedure of generating predicted pixel values of the parent population to generate such a way that the current generation function of the predicted values was included in the parent population.

13. Machine-readable storage medium on which is stored a program for encoding images according to which a computer performs a method of encoding images according to claim 1.

14. Machine-readable storage medium on which is stored a program for encoding images according to which a computer performs a method of encoding images according to claim 7.

15. Machine-readable storage medium on which is stored a program for decoding images, according to which a computer performs a method of decoding images according to claim 4.

16. Machine-readable storage medium on which is stored a program for decoding images, according to which a computer performs a method of decoding images of claim 10.



 

Same patents:

FIELD: information technologies.

SUBSTANCE: method for motion vector coding includes the following stages: selection of the first mode as the mode of information coding about a predictor of the motion vector in the current unit, and in this mode information is coded, which indicates the motion vector predictor at least from one motion vector predictor, or selection of the second mode, in which information is coded, which indicates generation of a motion vector predictor on the basis of units or pixels included into a pre-coded area adjacent to the current unit; determination of the motion vector predictor of the current unit in accordance with the selected mode, and coding of information on the motion vector predictor of the current unit; and coding of the vector of difference between the motion vector of the current unit and predictor of the motion vector of the current unit.

EFFECT: increased efficiency of coding and decoding of a motion vector.

15 cl, 19 dwg

FIELD: information technologies.

SUBSTANCE: share of cast combinations of optimal forecasting modes, which shall be selected for spatially corresponding units of upper and lower layers is identified on the basis of the optimal forecasting mode, which was selected in process of traditional coding, and a table of compliance is developed, which describes interconnections between them. Combinations of selected optimal forecasting modes in the compliance table are narrowed on the basis of the value of the share of casts, in order to create information of compliance for forecasting modes, which describes combinations of narrowed optimal forecasting modes. In process of upper layer unit coding, the version of searching for the forecasting mode, searching for which shall be carried out in process of coding, is identified by referral to information of compliance for forecasting modes using as the key the optimal forecasting mode selected in process of coding of the spatially corresponding unit of the lower layer.

EFFECT: reduced versions of searching for a forecasting mode of an upper layer using correlations of optimal forecasting modes between layers.

7 cl, 14 dwg

FIELD: information technology.

SUBSTANCE: displacement vectors are searched for by searching for global displacement, breaking up the image into multiple layers of blocks, successive processing of the layers using various search schemes, using displacement vector prediction, as well as selecting displacement vectors based on efficiency of their further entropy coding.

EFFECT: quality improvement of efficiency of a video compressing system, especially at low bit losses, high output thereof.

2 cl, 8 dwg

FIELD: information technologies.

SUBSTANCE: video coding device is a video coding device for exposure of a video image to forecasting coding with compensation of motion, comprising a detection module, in order to detect accessible blocks for blocks having vectors of motion, from coded blocks adjacent to a block to be coded, and a number of available blocks, a selection module, in order to select one selective block from coded accessible blocks, a coder of selection information, to code information of selection, indicating the selective block, using a coding table, corresponding to the number of accessible blocks, and a coder of images, to expose the block to be coded to forecasting coding with compensation of motion using a vector of motion of the selective block.

EFFECT: reduction of additional information by information of selection of a motion vector with increased extents of freedom for calculation of a motion vector by selection of one of coded blocks.

10 cl, 14 dwg

FIELD: information technology.

SUBSTANCE: each re-encoded frame of a multiview video sequence, defined according to a predetermined encoding sequence, is presented as a set of non-overlapping units; at least one of already encoded frame is determined, which corresponds to said view and denoted as reference; synthesised frames are generated for the encoded and reference frames, wherein for each non-overlapping unit of pixels of the encoded frame, denoted as the encoded unit, a spatially superimposed unit inside the synthesised frame is determined, which corresponds to the encoded frame, denoted as a virtual unit, for which the spatial position of the unit of pixels in the synthesised frame which corresponds to the reference frame is determined, so that the reference virtual unit thus determined is the most accurate numerical approximation of the virtual unit; for the determined reference virtual unit, the spatially superimposed unit which belongs to the reference frame, denoted as the reference unit, is determined, and the error between the virtual unit and the reference virtual unit is calculated, as well as the error between the reference virtual unit and the reference unit; the least among them is selected and based thereon, at least one differential encoding mode is determined, which indicates which of the units found at the previous should be used to perform prediction during the next differential encoding of the encoded unit, and differential encoding of the encoded unit is carried out in accordance with the selected differential encoding mode.

EFFECT: providing differential encoding of a frame using a small volume of service information by taking into account known spatial connections between neighbouring views at each moment in time, as well as information available during both encoding and decoding.

5 cl, 15 dwg

FIELD: information technology.

SUBSTANCE: method of encoding an image using intraframe prediction involves selecting a pixel value gradient which is indicated by the image signal to be predicted from among a plurality of selected gradients; generating a predicted signal by applying the gradient in accordance with the distance from the reference prediction pixel, based on the gradient; intraframe encoding of the image signal to be predicted, based on the predicted signal; and encoding information which indicates the value of the selected gradient. As an alternative, the method involves estimating the pixel value gradient which is indicated by the image signal to be predicted, based on the image signal already encoded; generating a predicted signal by applying the gradient in accordance with distance from the reference prediction pixel, based on the gradient; and intraframe encoding of the image signal to be predicted, based on the predicted signal.

EFFECT: improved image compression efficiency.

20 cl, 55 dwg

FIELD: information technology.

SUBSTANCE: method of encoding a video signal comprises steps of: forming a predicted image for the current block; generating a weighted prediction coefficient for scaling the predicted image; forming a weighted prediction image by multiplying the predicted image with the weighted prediction coefficient; generating a difference signal by subtracting the weighted prediction image from the current block; and encoding the difference signal, wherein generation of the weighted prediction coefficient involves calculating the weighted prediction coefficient for which the difference between the base layer image, which corresponds to the current block, and the predicted image is minimal.

EFFECT: high efficiency of encoding a video signal by reducing the error of the current block, which must be compressed, and the predicted image.

31 cl, 16 dwg

FIELD: information technology.

SUBSTANCE: deblocking filter 113 adjusts the value of disable_deblocking_filter-idc, slice_alpha_c0_offset_div2 or slice_beta_offset_div2 based on the Activity of an image calculated by an activity calculation unit 141, the total sum of orthogonal transformation coefficients of the image calculated by an orthogonal transformation unit 142, Complexity of the image calculated by the rate control unit 119, or the total sum of prediction errors of the image calculated by a prediction error addition unit 120.

EFFECT: improved image quality through correct deblocking.

8 cl, 7 dwg

FIELD: information technology.

SUBSTANCE: disclosed is an image decoding method comprising steps of parsing network abstraction layer (NAL) units of a base view (S200); decoding an image of the base view (S202); parsing multiview video coding (MVC) extension parameters of a non-base view (S204); searching whether or not prefix NAL units for a base view are present (S205); either calculating MVC extension parameters for the base view when no prefix NAL units are present (S206) or parsing the MVC extension parameters of the base view when prefix NAL units for the base view are present (S207); and decoding the non-base view using the MVC extension parameters of the base view and the MVC extension parameters of the non-base view (S210).

EFFECT: providing multiview video coding methods of multiview video decoding methods, even when prefix NAL units are not used.

2 cl, 23 dwg

FIELD: information technologies.

SUBSTANCE: method of video coding includes establishment of candidates of reference pixels for pixels within a previously specified range of distances measured from a target coding unit; generation of a predicted signal by means of serial selection of reference pixels used for inner prediction of the target coding unit, among reference pixels-candidates, whenever a condition of distance from the target coding unit varies, and by generation of a predicted signal by reference pixels for each condition of distance; calculation of costs for coding to implement coding with inner prediction of the target coding unit using each generated predicted signal; final detection of reference pixels used for inner prediction of the target coding unit, on the basis of each calculated cost for coding; and coding of information indicating position of detected reference pixels.

EFFECT: provision of efficient internal prediction of an image, which contains eclipses or noise, or to an image, where signals arise that have similar spatial frequencies, such images may not be processed by means of a regular internal prediction.

10 cl 18 dwg

FIELD: physics; video technology.

SUBSTANCE: invention relates to devices for re-encoding video data for real time streaming, and particularly to re-encoding video data for real time streaming in a mobile broadcast application. Proposed is a device for using content information to encode multimedia data, which includes a content classification module, which is configured to classify content multimedia data and provide content classification data, and an encoder which is configured to encode multimedia data in a first data group and a second data group based on content classification, wherein the first data group contains a coefficient, and the second group of data contains a first differential refinement associated with the coefficient of the first group of data.

EFFECT: design of a transcoder which provides for highly efficient processing and compression of multimedia data, which uses information defined from the said multimedia, and is scalable and error-tolerant for use in several multimedia data applications.

49 cl, 45 dwg

FIELD: technology for registration of digital information, possible use for increasing possibly recordable volume on DVD or CD.

SUBSTANCE: method generates floating compression coefficient ranging from 2 to 255 during recording by comparing values of codes of stream based on value, counting number of equal and following each other, codes, generation of binary code of this number and insertion of it into stream following first code of its series while excluding counted codes from it, during restoration number of equal codes is determined for stream, decrypted and number of signals for outputting first code is generated, equal to number of removed codes during compression.

EFFECT: increased compression level during recording and improved digital data restoration quality together with excessive information.

2 cl, 4 dwg

The invention relates to automation and computer engineering and can be used in computers to convert the p codes Fibonacci unitary code

The code converter // 2023347
The invention relates to automation and computer engineering and can be used in computers to convert the numbers from the lowest form of redundant number systems

FIELD: technology for registration of digital information, possible use for increasing possibly recordable volume on DVD or CD.

SUBSTANCE: method generates floating compression coefficient ranging from 2 to 255 during recording by comparing values of codes of stream based on value, counting number of equal and following each other, codes, generation of binary code of this number and insertion of it into stream following first code of its series while excluding counted codes from it, during restoration number of equal codes is determined for stream, decrypted and number of signals for outputting first code is generated, equal to number of removed codes during compression.

EFFECT: increased compression level during recording and improved digital data restoration quality together with excessive information.

2 cl, 4 dwg

FIELD: physics; video technology.

SUBSTANCE: invention relates to devices for re-encoding video data for real time streaming, and particularly to re-encoding video data for real time streaming in a mobile broadcast application. Proposed is a device for using content information to encode multimedia data, which includes a content classification module, which is configured to classify content multimedia data and provide content classification data, and an encoder which is configured to encode multimedia data in a first data group and a second data group based on content classification, wherein the first data group contains a coefficient, and the second group of data contains a first differential refinement associated with the coefficient of the first group of data.

EFFECT: design of a transcoder which provides for highly efficient processing and compression of multimedia data, which uses information defined from the said multimedia, and is scalable and error-tolerant for use in several multimedia data applications.

49 cl, 45 dwg

FIELD: information technology.

SUBSTANCE: disclosed is use of a parent population which is generated via random formation of a procedure for generating a predicted value, each indicated by a tree structure, and a set of procedures for generating a predicted value is selected as a parent from such a population. The procedure for generating a predicted value is generated as a descendant based on a certain method of development of the tree structure which develops selected procedures for generating a predicted value, where the existing function for generating a predicted value can be a tree end node. The procedure for generating a predicted value, having the best estimate cost, is selected from procedures for generating a predicted value as a parent and a descendant, and overall information content for representing the tree structure and volume of the code, estimated by the predicted pixel value, is used as a cost estimate, and the final procedure for generating a predicted value is formed by repeating the relevant operation.

EFFECT: high encoding efficiency.

28 cl, 14 dwg

FIELD: physics.

SUBSTANCE: code size control method used in the method of video coding for execution of the code size control by assessment of the code size generated for a coded target frame, meanwhile the control method includes: the phase, during which the reference value of the coded target frame is evaluated and this value is saved in the storage device; the phase, during which the reference value of earlier coded frame is retrieved, which is used for assessment of the generated code size; the phase, during which the reference value of the coded target frame is compared with the reference value of the earlier coded frame; And the phase, which is performed according to the result of matching of reference values, during which, if is established that the difference between both reference values is greater than the earlier set criterion value, and the coded target frame is more complex, than the earlier coded frame, the code size generated for the coded target frame is estimated, without using the coding result of the earlier coded frame, and otherwise at this phase the code size generated for the coded target frame is estimated, on the basis of result of coding of the earlier coded frame.

EFFECT: preventing of deterioration of image quality.

9 cl, 5 dwg

FIELD: information technology.

SUBSTANCE: codec (30) includes at least one coder (10) and at least one decoder (20). Encoder includes data processing circuit for application to input data (D1) of one of forms of differential and/or summing coding to form one or more corresponding coded sequences, which is subjected to cyclic shift relative to maximum value and/or cyclic shift relative to minimum value to generate encoded output data (D2 or D3). Decoder includes data processing circuit for processing one or more parts of encoded data (D2 or D3) configured to use one of difference and/or summing decoding types to one or more corresponding coded sequences specified one or more parts, wherein one or more encoded sequences are subjected to cyclic transition operation relative to maximum value and/or cyclic transition relative to minimum value for formation of decoded output data (D5).

EFFECT: high degree of data compression.

44 cl, 3 dwg, 2 tbl

FIELD: video decoders; measurement engineering; TV communication.

SUBSTANCE: values of motion vectors of blocks are determined which blocks are adjacent with block where the motion vector should be determined. On the base of determined values of motion vectors of adjacent blocks, the range of search of motion vector for specified block is determined. Complexity of evaluation can be reduced significantly without making efficiency of compression lower.

EFFECT: reduced complexity of determination.

7 cl, 2 dwg

FIELD: compensation of movement in video encoding, namely, method for encoding coefficients of interpolation filters used for restoring pixel values of image in video encoders and video decoders with compensated movement.

SUBSTANCE: in video decoder system for encoding a video series, containing a series of video frames, each one of which has a matrix of pixel values, interpolation filter is determined to restore pixel values during decoding. System encodes interpolation filter coefficients differentially relatively to given base filter, to produce a set of difference values. Because coefficients of base filter are known to both encoder and decoder and may be statistically acceptably close to real filters, used in video series, decoder may restore pixel values on basis of a set of difference values.

EFFECT: efficient encoding of values of coefficients of adaptive interpolation filters and ensured resistance to errors of bit stream of encoded data.

5 cl, 17 dwg

Up!