Method of adaptive quantization vectors

 

(57) Abstract:

The invention relates to the field of digital signal processing image and can be used when transmitting video over narrowband communication channels and implementing quick view large archival databases. The method uses the adaptive quantization vectors. Preserving the topology of the image compression is achieved by the fact that each of the vectors is determined by the minimum Euclidean distance with respect to the reference. Decompression is carried out in relation to the selected reference points in the image.

The invention relates to the field of signal processing image and can be used in the processing and archiving large amounts of video data. For example, when creating a large archive of graphic databases, the implementation of procedures quick view, including encyclopedias on CD-ROM, and when viewing and processing large quantities at a time and at a high tempo information.

Known technical solution for U.S. patent N 5157488, H 04 N 7/13, 1994 u. "Method for adaptive quantization in progressive mode", namely, that when the adaptive quantization in the transmission when Mantovani for each component image. For some tables of quantization of the components of the image determine the overall conversion rate and transmit signals measurement of sequential blocks of video data by passing the General conversion factor as the next component together with components of the image and, thus, perform adaptive quantization video.

The disadvantage of this method of adaptive quantization of video data is the presence of boundaries of groups of data in the restored image when the ratio of > 10.

There is a method of video compression digital television channels (patent of Russian Federation N 2017347 H O 4 N 7/13, 1994), namely, that in the transmit channel delay on the transmission time of 1 picture element, subtract from the direct video signal delayed video signal in each of the fifth picture element in the full code coded signal is a reference signal, to form the inter-element signal-difference simultaneously with the formation of the reference signal and the signal interelement difference, forming a signal line of the differential path delay of the direct signal on one line and subtracted from the direct signal interelement difference in Solow line and inter-element differences, the small signal components are passed alternately with a reference signal in the full code, summed up in odd rows of the reference signals with signals interelement difference, and in even-numbered rows summarize the reference signals of the odd row signal line difference, restored the reference signals are also summed up in even rows with signals interelement difference.

However, compressing the image at the specified method with good quality can only be achieved for simple images and requires a large hardware cost, in addition, is not stored topological structure of the original image.

The invention solves the problem of creating effective method of image compression on the basis of classical methods adaptive quantization vectors, preserving the topology of the original image.

The closest analogue of the claimed technology solutions is a device for encoding and decoding using vector quantization for U.S. patent N 5010401 H O 4 N 7/13, 1991

Converting the image signal at the specified patent is as follows. The signal is divided into blocks according to the counts in each (positive integer) are formed blocks mikekoenigs of image blocks, the code block of the image determines the address of the block in the library of the standards when restoring read this block in the buffer memory with the subsequent formation of the reconstructed image.

The classical method of adaptive quantization vectors is based on preliminary building code tables (code libraries of standards-based learning (adaptation) compression specific blocks of the image.

Used libraries do not preserve the topology of the encoded data, resulting in large errors when encoding in terms of interference. The proposed method of image compression saves the specified topology due to optimization of the learning process and flow as follows.

The original image is divided into blocks of pixels of a certain size, each of which is interpreted as a vector in n-dimensional space. Method adaptive clustering divide this space into m clusters such that density of the clusters corresponded to the probability density vectors in the specified vector space.

Before adaptive clustering designate centers casanellas sampling the training data (n-dimensional vectors), i.e., all blocks of pixels of the encoded image. Based on the analysis of these samples is the adaptation of the provisions of these centers to the statistical properties of the encoded data. For each of the vectors is determined by the deviation from the centers mentioned clusters according to the criterion of the minimum of the Euclidean distances. For each cluster a new center position is calculated as the average of all those caught in the cluster vectors.

In the procedure of training blocks in the code table are ordered by all visual characteristics. Therefore, when performing the encoding operation (replacement of blocks of the image numbers from the code library) General view of the image, i.e., its topology is preserved.

In the encoding process by determining the smallest of the Euclidean distances between the current vector and reference each encoded block of pixels is in accordance "to the closest reference vector from a table encoding.

For transmission over the communication channel or backup compressed image can be decoded. The value of each pixel of the decoded image is the address of the entry in the decoding table, in which joblocom 4x4 pixel. Another important distinguishing feature of the proposed method is that in accordance with the selection of the reference image pixel values of the reference pixels is converted into values of the reference vectors. In accordance with the selected control points in the decompression on the screen of the selected area of the image, i.e., implemented direct access to any portion of an image. This allows to considerably reduce the time for reading the image and outputting it to the screen.

Since the compressed image look like the original, it allows when working with the video data, some preliminary operations, in particular, can be pre-bind image to predetermined standard. This significantly reduced the binding.

Thus, the proposed method has the following distinctive features. When preparing a code library blocks are ordered according to their visual characteristics, so small errors in the selection of blocks in the encoding process or transfer does not result in a significant loss in the quality of the reconstructed image, thus providing povysheniya topological structure of the original image. That is, an array of compressed image when interpreting the code blocks as brightness, looks outwardly similar to the original image (of course, losing spatial resolution of the original image is proportional to the used compression ratio). This feature allows to receive additional benefits in the practical application of the method of image compression:

1. Direct access to compressed images.

In this case, to ensure more rapid preliminary analysis of the image, the user can view the compressed image directly without decompression, to produce a preliminary analysis of the image. After selecting the desired plot is decompression on the screen (or in memory) of the given image. That is implemented direct access to any portion of an image. This allows to considerably reduce the time for reading out the image from the computer's memory (as read compressed many times, instead of the full image, save the image on the screen (as when viewing compressed image smaller than the original).

Compressed images can find its primary use in military applications when creating a large archive of graphic databases (with implementation procedures quick view), when viewing and processing large quantities at a time and in a high rate of video information, for example, aerospace.

1. Method of adaptive quantization vectors, namely, that the image is divided into blocks of pixels of a certain size, each of which is interpreted as a vector in n-dimensional space, creating quantieme vector prostranstvennoi and the squared difference between the current signal value and the corresponding library standards, pairwise sum of the squared differences produce a total memorize the received code block from the buffer memory in accordance with its address in the library standards and form a reconstructed image, wherein the local block from a library of patterns corresponding to each block of the image is found by determining the minimum of the Euclidean distances between the current vector and the reference, select the reference point on the original image, recalculate their values in the value of the reference vectors, the decompression is carried out only certain parts of an image, defined anchor points.

2. The method according to p. 1, characterized in that prior to decompression are binding image to predetermined standard.

 

Same patents:

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

FIELD: video encoding, in particular, methods and devices for ensuring improved encoding and/or prediction methods related to various types of video data.

SUBSTANCE: the method is claimed for usage during encoding of video data in video encoder, containing realization of solution for predicting space/time movement vector for at least one direct mode macro-block in B-image, and signaling of information of space/time movement vector prediction solution for at least one direct mode macro-block in the header, which includes header information for a set of macro-blocks in B-image, where signaling of aforementioned information of space/time movement vector prediction solution in the header transfers a space/time movement vector prediction solution into video decoder for at least one direct mode macro-block in B-image.

EFFECT: creation of improved encoding method, which is capable of supporting newest models and usage modes of bi-directional predictable (B) images in a series of video data with usage of spatial prediction or time distance.

2 cl, 17 dwg

FIELD: movement estimation, in particular, estimation of movement on block basis in video image compression application.

SUBSTANCE: method and device are claimed for conducting search for movement in video encoder system using movement vectors which represent difference between coordinates of macro-block of data in current frame of video data and coordinates of corresponding macro-block of data in standard frame of video data. A set of movement vector prediction parameters is received, where movement vector prediction parameters represent approximations of possible movement vectors for current macro-block, movement vector search pattern is determined and search is conducted around each movement vector prediction parameter from the set of movement vector prediction parameters using search pattern, and on basis of search result, the final movement vector is determined.

EFFECT: increased efficiency of video signals compression.

3 cl, 7 dwg

FIELD: physics.

SUBSTANCE: said utility invention relates to video encoders and, in particular, to the use of adaptive weighing of reference images in video encoders. A video encoder and a method of video signal data processing for an image block and the specific reference image index for predicting this image block are proposed, which use the adaptive weighing of reference images to increase the video signal compression, the encoder having a reference image weighting factor assigning module for the assignment of the weighting factor corresponding to the said specific reference image index.

EFFECT: increased efficiency of reference image predicting.

8 cl, 7 dwg

FIELD: physics, computing.

SUBSTANCE: invention relates to the field of coding and decoding of a moving image. In the method, at least one reference image for the processing of the field macroblock is selected from at least one reference image list, using information about reference image indexes, each at least one reference image selected is a field, and the parity of at least one reference field selected may be based on the parity of the field macroblock and the reference image index information.

EFFECT: efficient provision of information about reference image compensating motion, by reference image indexes determined in different ways, according to the coded macroblock modes.

10 cl, 12 dwg

FIELD: information systems.

SUBSTANCE: invention refers to video coders using adaptive weighing of master images. The video decoder for decoding data from video signal for the image, having multiple motion boxes, containing: the master image weighting coefficient module for accepting, at least, one master image index, thereat each one from the mentioned master image indexes is intended for independent indication not using any other indexes, one of the multiple master images, used for prediction of current motion box and weighting coefficient from the set of weighting coefficients for current mentioned one from mentioned multiple motion boxes.

EFFECT: increase of efficiency in predicting master images.

20 cl, 7 dwg

FIELD: information technology.

SUBSTANCE: method is offered to compress digital motion pictures or videosignals on the basis of superabundant basic transformation using modified algorithm of balance search. The algorithm of residual energy segmentation is used to receive an original assessment of high energy areas shape and location in the residual image. The algorithm of gradual removal is used to decrease the number of balance assessments during the process of balance search. The algorithm of residual energy segmentation and algorithm of gradual removal increase encoding speed to find a balanced basis from the previously defined dictionary of the superabundant basis. The three parameters of the balanced combination form an image element, which is defined by the dictionary index and the status of the basis selected, as well as scalar product of selected basic combination and the residual signal.

EFFECT: creation of simple, yet effective method and device to perform frame-accurate encoding of residual movement on the basis of superabundant basic transformation for video compressing.

10 cl, 15 dwg

FIELD: information technology.

SUBSTANCE: playback with variable speed is performed without picture quality deterioration. Controller 425 creates EP_map () with RAPI address in videoclip information file, dedicated information selection module 423 RAPI, image PTS with internal encoding, which is immediately preceded by RAPI, one of final positions of the picture with internal encoding, as well as the second, the third and the fourth reference pictures, which are preceded by the picture with internal encoding. The controller saves EP_map () in output server 426, i.e. controller 425 copies the value, close to given number of sectors (quantity of sectors, which can be read at one time during encoding process) of final positions for the four reference pictures (1stRef_picture, 2ndRef_picture, 3rdRef_picture and 4thRef_picture) to N-th_Ref_picture_copy, defines value of index_minus 1 on the basis of N-th_Ref_picture_copy and records it to disc.

EFFECT: effective process performance with constant data reading time.

8 cl, 68 dwg

FIELD: information technology.

SUBSTANCE: invention proposed contains videodecoder (200) and corresponding methods of videosignal data processing for image block with two reference frames' indices to predict this image block. The methods use latent scaling of reference images to improve video compressing. The decoder (200) contains latent scaling coefficient module (280) of reference images, which are used to determine a scaling coefficient value, corresponding to each of the reference image indices. Decoding operations contain receiving reference image indices with data, which corresponds to image block, calculation of latent scaling coefficient in response to image block location relative to reference images, indicated by each index of reference image, extraction of reference image for each of the indices, motion compensation relative to extracted reference image and multiplication of reference images, relative to which the motion compensation was performed, to a corresponding scaling value.

EFFECT: increase of decoding efficiency.

25 cl, 6 dwg

FIELD: digital processing of images, possible use for transmitting images through low speed communication channels.

SUBSTANCE: in accordance to the invention, the image is divided onto rank blocks, for each rank block of original image a domain or a block is found in the code book and a corresponding transformation, which best covers the given rank block, if no sufficiently precise match is found, then rank blocks are divided onto blocks of smaller size, continuing the process, until acceptable match is achieved, or the size of rank blocks reaches certain predetermined limit, while after the division of the image onto rank blocks, classification of the blocks is performed, in accordance to which each domain is related to one of three classes, also except classification of domain blocks of original image, code book blocks classification is also performed, and further domain-rank matching is only performed for those domains, which belong to similarity class of given rank area. As a result, during the encoding, the search for area, which is similar to a rank block, is performed not only among the domains which are blocks of the image being encoded, but also among the code book blocks which match the rank area class.

EFFECT: increased speed of encoding with preserved speed of transmission and frame format length.

3 dwg

Up!