Controlling speed of scalable coded images

FIELD: information technology.

SUBSTANCE: subsets are determined (step 29), each containing one or more coding units, where at least one image puts at least one coding unit into two or more subsets, the list of requirements (LOR) is established (step 30) containing at least one element associated with each subset. Significance values are use in order to select quality increments for generating an allowable code stream which satisfies the LOR for subsets (steps 34, 36). Quality increments can be selected so as to attain high quality for different subsets depending on size requirements in the LOR. For certain requirements, the code stream will exhibit an approximately constant quality of the reconstructed image. Quality increments can be selected so as to achieve small sizes of a compressed image for different subsets depending on quality requirements in the LOR.

EFFECT: high quality of the reconstructed image.

27 cl, 7 dwg

 

The technical field to which the invention relates.

The invention generally relates to a method of speed control for scalable coded images and, more particularly, to a speed control that satisfies the given requirements for subsets of the image data.

Description of the prior art,

Over the last few decades encodes a frequency band or wavelet coding has confirmed that it is an effective way to compress images. Particular importance is a new compression standard JPEG2000 image described in the recommendations of the sector of International telecommunications Union (ITU-T) ITU-T Rec. T.800/ISO/IEC 15444-1:2004 JPEG 2000 Image Coding System for image coding JPEG 2000), thereby fully incorporated herein by reference. Like other compression standards, the JPEG2000 standard defines the decoder and associated syntax of the code stream. The standard does not prescribe the actions of the encoder, as the generated code stream is subject to a specified syntax code stream and can be decoded by the corresponding decoder. This enables the development of adaptable encoder. Cm. the publication "JPEG2000 Image Coding System," 2004 and D.S. Taubman and M.W. Marcellin, JPEG2000: Image Compression Fundamentals, Practice and Standards (JPEG2000: basic principles, p is actica and standards for image compression), published in Kluwer Academic Publishers,Boston, 2002, which thereby are fully incorporated herein by reference.

Figure 1 illustrates a typical JPEG2000 encoder 10 that is used to encode the image 11. Each image (optionally) is divided into non-overlapping rectangular region of cells 12. Region-cells enable random access to space and limit the memory requirements during implementation. You can then use the optional conversion component 14, in order to improve the compression efficiency. For example, if the image consists of the color components "red", "green" and "blue", the use of color conversion can improve the performance of the compression. Each (transformed) color component region-cells hereinafter referred to as component-area. The application of wavelet transform 16 to each component area creates a set of transform coefficients arranged in the frequency ranges for each component area. Conversion factors for each frequency band are then divided into rectangular blocks, called to be coding blocks 18 coefficients. Each encode a block of coefficients are then coded independently by the encoder 19, the block is at odds.

For a given block of coefficients implementation of its encoding begins by quantization its coefficients to obtain the quantization indexes. These indexes quantization can be considered as an array of integers. When using a reversible wavelet transform, quantization is not strictly required, since the wavelet coefficients are integers. Such an array of integers can be represented using character array and array values. The character array can be considered as a binary array, where the array value at each point indicates whether the index of the quantization positive or negative. Array values can be divided into a number of binary arrays with one bit of the index index of quantization. The first of these arrays corresponds to a senior significant bits (MSB) of the quantization indices, and the last corresponds to the youngest significant bits (LSB). Each array is called a bit plane. Each bit plane of the block coefficients are then entropy encode using encoder bit plane. Used in JPEG2000 encoder bit plane is context-dependent, binary arithmetic coder. Encoder bit plane performs three passes for each bit plane of the block of coefficients. These PR the passages referred to code passages. Each bit in the bit plane is encoded in one of these code passes. The resulting compressed data is called compressed code passes.

The encoder block of coefficients calculates the magnitude of the decrease distortion (RMS error), provide each compressed code passage, along with the length of the compressed code of the passage. With this information it is possible to ask for the reduction of distortion along the length of the compressed code of the passage in the form of a slope index corresponding to the intensity of the distortion of the compressed code of the passage. The slope index corresponding to the intensity of the distortion, and the compressed code of the passage is attributable to one byte with value of reduction of distortion provided by the compressed code pass. Thus, the compressed code pass with a higher slope corresponding to the intensity of the distortion can be considered more important than one with a smaller slope index corresponding to the intensity of the distortion. The encoder 19 blocks of coefficients delivers the compressed code passages 20, their lengths 22 and indicators 21 of inclination corresponding to the intensity of the distortion, block 23 generating code stream, which decides which of the compressed code of the passages 20 of each block 18 factors will be included in a code is Otok. The block generation code stream includes a code stream of compressed code passages with the greatest indicators of inclination corresponding to the intensity of the distortion until exhausted budget (resource) bytes.

If JPEG2000 is used for compression of image sequences, there are only a few currently known ways to determine what speed to use for each image of the sequence. One possibility is to select a fixed speed (i.e., a fixed number of bytes)to encode each image of the sequence. Although this method is simple and allows for a simple implementation, it does not give adequate performance in some applications. In many image sequences characteristics of the images in the sequence to change significantly. Because this method specifies a fixed number of bytes for each image, the resulting sequence decompencirovannah images characterized by large changes in the quality between images.

This deficiency was identified Tzannes, etc. in the application for U.S. patent 2004/0047511 A1. To achieve a slight improvement in performance Tzannes, etc. make it possible adaptive choice of compression options when image coderoute the series. Adaptation is performed for the current image using the information collected only on the basis of previous images in the sequence: the following images are not considered in the allocation rate for the current image. In addition, if two consecutive images in the sequence are not vysokokoncentrirovannym (for example, as in the case during the scene), adaptation becomes unstable. Another alternative to coding with fixed speed was presented Dagher and others in the Institute's materials engineers electrical and electronics (IEEE) Resource-Constrained Rate Control for Motion JPEG2000 (speed control transmission of moving images with limited resources), Transactions on Image Processing, December 2003. In this way the compressed image is placed in the buffer. The compressed data extracted from the buffer at a constant speed. New compressed image is added to the buffer as they become available. If the buffer is full, when it should add a new compressed image, this new compressed image, and other images already in the buffer, truncate, so that all compressed data will fit in the buffer. The resulting images have a relatively low change in quality within a time window of variable duration, is suitable is th the length of the buffer. However, the quality may vary significantly in the time frames longer than the window of variable length.

In addition, none of the above methods provides the ability to define requirements for size or quality for subsets of the image data, such as individual images, the individual components etc.

A brief description of the invention

The present invention provides a method of speed control for the sequence of scalable coded images, which satisfies the requirements that apply to subsets of the image data.

The method of speed control is applicable to the class of coders, in which image in the sequence is converted to obtain transform coefficients. These factors are divided into blocks of coding. The factors for each coding block is encoded, for getting multiple increments quality. In the system of JPEG2000 compression units of encoding are subject to encoding blocks of coefficients and increments are compressed code passes.

The speed control is performed by collecting blocks of encoding one or more subsets. Associated with the image subset is composed of the blocks of the coding of the individual images. the La given image subset, associated with the image, can set different levels of permissions that region of space, the field-cells, color components, or any combination thereof, including the full image. Associated with a sequence of subsets composed of to be encoded blocks of coefficients from each image in the sequence. They can ask for a full sequence of different levels of resolution, areas in space, area, cell, or color components, or any combination of them. A single associated with the sequence of the subset may include all blocks of the coding for all images in the sequence, determining as a consequence, the sequence of the full image.

The list of requirements (LOR) create so that each subset had a list of at least one associated element. Requirements relating to the subsets associated with the image, referred to as associated with the image requirements. Requirements relating to the subsets associated with the sequence, referred to as associated with the sequence requirements.

The values of significance are calculated for each increment of quality for each coding block. The values of significance are used to determine which code passes save to udovletvor the ü the requirements of the relevant subsets when trying to achieve high quality remanufactured or low sizes of the compressed image. The code stream that meets the LOR for all subsets, create a sequence of images based on the stored increments quality. Such a code is called "valid code stream. The way to determine which increment quality save for each coding block, and create a valid code flow will depend on such aspects as LOR, computational complexity, memory resources, etc.

The invention briefly described for the case of attempts to achieve a high quality, according to the specified size requirements. The achievement of low size, as permitted according to the specified quality requirements, is similar. For each image in a single-pass approach selected increment of quality, with the highest values of significance within each associated with an image subset, so that the total size of the selected increments quality satisfies associated with the image the size requirements for this subset. The selected increment of quality, value relevance and length are stored. The remainder of the cast. Once all the images are processed in a similar manner, the remaining increment of quality for each associated with a sequence of subsets that have the highest value of importance, are chosen so that the total size of the SEL is data increments quality met-related sequence the size requirements for this subset. Selected increment quality form a valid code stream, and the remainder is discarded.

Each method strives to achieve high quality of the reconstructed image within the subsets, as permitted according to the requirements. Specifically, methods are making efforts to achieve high quality within each associated with an image subsets for each individual image, as well as the high average quality (averaged over all images or sequence) within each associated with a sequence of subsets. The problem of compromise between the high quality from image to image and high average quality may be due to the relative stringency of requirements to the size associated with the image, against those associated with the sequence. When over the choice of the increment of the quality dominate any related sequence size requirements, each of the methods will also be approximately constant quality of the reconstructed image from image to image within the respective subsets. These and other features and advantages of the invention will be obvious to experts in the art from the following detailed descriptions of preferred options to implement the Oia, taken together with the accompanying drawings, in which:

Figure 1 - block diagram of a typical JPEG2000-encoder as described above;

Figure 2 - block diagram of the sequence of operations illustrating the allocation rate for the sequence of images in accordance with the present invention;

Figure 3 - illustration of the selection of the increments of the quality of the subset;

Figure 4 - illustration of an exemplary organization code stream when there are multi-component image;

Figure 5 - illustration of the wavelet sub-bands, which contribute to different resolutions for multi-component images;

6 is a chart of the sequence of actions that illustrates the integration of speed control in the process flow of processing digital image (DI); and

7 is a schematic diagram of the system of the encoder workstation DI and speed controller.

Detailed description of the invention

The invention generally relates to a method for encoding images using class scalable coders, shown, for example, JPEG2000, and, more specifically, to a method of speed control for the sequence of scalable coded images. The way the speed control determines the compressed data to be included in the valid code stream for a sequence of images so that in the established set of requirements has been satisfied for a predetermined definition of the subset. According to the method, an attempt is made to achieve high quality of the reconstructed image within the respective subsets in accordance with the assumption according to the size requirements for subsets. For some requirements, the method will also be approximately constant quality of the reconstructed image from image to image within the respective subsets. Of course, limited by the small size of the subset will be low as compared with that if it was limited to larger size. Thus, the "high quality" should be interpreted relative to requirements. Alternatively, according to the method may attempt to reach a small compressed size for subsets satisfying the specified quality requirements for subsets.

Describes how to control the rate applicable to any sequence of digital images, including movies, video, time-series data, three-dimensional image (such as multispectral and hyperspectral data, remote sensing images), or data sets of higher dimension, and a single static image. While the embodiments of described regarding the JPEG2000 compression, they are applicable to the other compression methods with functionality, similar to that of JPEG2000.

Although some embodiments of the present invention is described relative to a single image sequence, they are also applicable for cases where the image sequence consists of many groups. As an example, such groups can be obtained by time-division sequence. In this case, each group could be similar to the "part" of the movie. As another example, different groups can be set in view of the characteristics, the end portion 1, the end portion 2, commercials, etc. In the case of image sequence 3D(stereo) can be two groups, consisting of images intended for viewing the left and right eyes, respectively. "Left" and "right-wing" groups can be further broken down into "parts". In each case, the selection of the transmission speed can be performed separately on each group by considering each as a sequence. On the other hand, all groups can be aggregated to identify the speed to be treated as a single sequence.

Speed control for scalable coded images

As shown in figure 2, the speed control performed by identifying subsets of coded blocks is I (step 29). Preferably the units of encoding are to be encoded blocks of coefficients in the JPEG2000 standard, but alternatively, frequency bands, blocks of coefficients of discrete cosine transform (DCT), the transformed components, full transformed image, or other data structures. These subsets are determined with the help of information about the encoding process, which will be used on the stage 31.

The subset defined in relation to individual images, referred to as associated with the image subset. These are associated with the image subset can set different levels of permissions, for example 2KB and 4Kb, different areas in space, such as foreground and background, different color components such as R, G, B, or combinations thereof, including the full image. Are other subsets within the essence and scope of the method. Associated with the image of the subset is useful in determining the requirements for individual images. For example, in multi-component application requirements to the size of the compressed image may be imposed on each component of the image and full image for three-component image blocks, the image encoding will then be grouped in three subsets of the, corresponding to the three components of the image, for example, R, G, B. the Fourth subset is composed of all the coding blocks in the image. It is obvious that for image-related subsets need not be disjoint.

The subset defined for the complete sequence of images, called the associated sequence subset. Associated with a sequence of subsets can set different levels of permissions, for example 2KB and 4Kb, different areas in space, such as foreground and background, or different color components such as R, G, B, or combinations thereof, including the full sequence. Other related sequence subsets are possible within the essence and scope of the method. Associated with a sequence of subsets are useful in determining the aggregate requirements for the full sequence. For example, in multi-component application requirements may be imposed on the total compressed size (aggregated across all images in the sequence) for each component, as well as the total compressed size for all images in the sequence. For a sequence of images, with each of the three components, blocks encode all images in the sequence bootsgenuine in three subsets, the corresponding R (red), G (green), B (blue). The fourth subset is composed of all the coding blocks all images in the sequence. It is clear then that associated with a sequence of subsets need not be disjoint.

Associated with the image subset for one image may differ from those of another image. For example, one image may be three subsets corresponding to R, G, B, while the other image has a single subset, corresponding to the full image. Another image may not have any defined subsets. Similarly, the sequence may not have any associated with the sequence defined subsets. At least one image must make one or more blocks encoding the two or more subsets. One image can make the same blocks coding in two subsets, or it can make various blocks of the coding in two subsets. Two subsets can be both related to the image, or both associated with a sequence, or one of each.

The list of requirements (LOR) is set for subsets (step 30), so that each subset has at least one requirement. LOR typically may include the step part of the quality requirements (required, minimum, maximum) and/or size requirements (required, minimum, maximum) for each of the subsets. Other requirements are possible within the essence and scope of the method.

The speed control is applied to each image, and an image sequence. To this end, each image is encoded using the encoder, such as JPEG2000 (step 31). Although the encoding process entails a number of stages that are understandable to experts in the field of engineering, basic and essential steps for the purposes of speed control are: (1) convert each image into transform coefficients using a wavelet transform, DCT or other suitable conversion; (2) separation of the transformed coefficients into blocks of coding and (3) encoding of the coefficients of each coding block to generate a lot of increments quality for each coding block. JPEG2000 uses wavelet transform divides the coefficients into blocks of coefficients that are coded to give a short code passes for each block of coefficients.

In JPEG2000 encoding units encoding applies quantization, followed by encoding the bit plane, to produce the increment of quality. This viewpoint systems is no quantization as a part of the process of encoding units encoding. Alternatively, the quantization can be considered as part of the conversion process. In fact, a particular implementation may include quantization in the transform through the use of appropriate scaling and rounding and/or reversible transformation. Usually the size of the quantization steps set relatively small, and all bit planes are encoded to provide high basic quality first allocation speed. Alternatively, a more moderate step size of quantization can be used to reduce the initial number of bit planes, which can be coded, resulting in a lower initial level of quality. Alternatively, it can be coded reduced number (preferable all of them) the most significant bit planes in order to achieve a similar effect. The reasoning in support of these alternatives is to save on complexity and/or memory/external memory. However, it should be paid attention to is not too much to reduce the initial quality. Alternatively, the size of the quantization step can be selected in accordance with the required quality.

Calculate (step 32) the values of significance for each increment of quality. This may be done simultaneously with step 31. P is chodashim values of significance may be indicators of slope, the corresponding intensity distortions. Such indicators tilt represent the ratio benefits/costs to enable increment of quality in the code stream. Specifically, the slope index for a particular increment of quality is the reduction in the distortion (usually root mean square error (MSE) or some other indicator of the quality represented by the increment value quality, divided by the length increment of quality (usually in bytes). Increment quality with larger indices of inclination corresponding to the intensity of the distortion, and then can be considered more important than those with smaller values of slope. As stated above, calculate the MSE can be modified to include consideration of visual weighting and/or visual masking (see, for example, Taubman and Marcellin, Section 16.1). In the alternative, the MSE can be replaced only significant difference" or other visually motivated distortion measures.

Are additional possible choices for the values of significance. For example, it can be assumed that the units coding for a subset have each K bit planes, numbered from K-1 to 0 (from MSB to LSB). These numbers of bit planes can be used as values of significance. A similar effect can be obtained by the use of the Finance of rooms increments quality (preferably, than numbers of bit planes) as the value of significance. Another variation is to assign a range of values of significance for all increments quality that have the same number of bit planes. Ranges for different numbers of bit planes will be different and the values of significance within the range can be ordered in accordance with the slope index corresponding to the intensity of the distortion. For example, let b be the number of bit planes for the increment of the quality and let s be its slope index corresponding to the intensity of the distortion. Additionally, let ms be the maximal slope index. Then one choice for the values of significance of the increment may be b+s/ms.

The values of significance can be weighted to select or deselect one or more fields in space (possibly the whole image) of one or more images.

As is apparent from Section 8.2 Taubman and Marcellin, sometimes it is desirable to prevent the end of the code stream of the coding block between some increments quality. Equivalent may require a "group" two or more increments quality and treat them essentially as a single composite increment of quality with the aim of identifying speed. This compound paradisiacally has a single slope index, the corresponding intensity of distortion calculated in the form of a General reduction in distortion for a group of increments quality divided by the total length of the group increments quality. To simplify the discussion, we should put in all respects, that this grouping is performed, when it occurred, and that then the term "increment quality" may refer to a compound increment.

The values of significance are used to determine the increment of quality save if you want to satisfy some requirements to a subset (step 34). The stages 31, 32 and 34 are performed for each image in the sequence (step 35).

It may seem that the drawing indicates the sequential processing of images, but parallel processing is also possible, since there is no dependency between images on the stages 31, 32 and 34. At this point a valid code flow for the full sequence is created with regard to any remaining requirements for the respective subsets (step 36).

In an exemplary embodiment, any associated with the image requirements are satisfied at step 34, while any associated sequence requirements meet on the stage 36. Alternatively, it is possible to save all the increments quality stage 34 and delay all however the decision to stage 36. This will increase the required RAM/external memory, but may have advantages in some applications, such as editing and creating archives.

The way to determine which increment quality save for each encoded image (step 34), and create a valid code stream for a sequence of images (step 36) will depend on the requirements and any additional issues such as computational complexity, memory, decoder, etc.

First describes the options for implementation, which attempts to achieve high quality, in accordance with the assumptions according to the size requirements. Subsequently will be described embodiments of which are making efforts to achieve small size, permissible according to the quality requirements.

Single-pass approach

Single-pass approach selects from each associated with an image subset increment of quality, with the highest value of importance, so that the total size of the selected increment of the quality of the subset satisfy any related image to your requirement size for this subset (step 34).

The selected increment of quality, values of significance and retain length for each image. N is selected increment quality reject. Once all images processed in this manner, the method selects the increment of the quality of all images), with the highest value of importance of those remaining in each linked with the image of the subset, so that the total size of the selected increment of the quality of that subset (in the aggregate for all images) satisfy any associated with the sequence requirement size for a subset (step 41). The selected increment of quality are used to generate valid code stream (step 42). The residue (unselected increments quality) is discarded and a valid coded stream recorded on a storage medium such as a storage drive, disk or tape, or it is transmitted over the channel.

Two pass approach

Single-pass approach may require temporary storage of compressed data, which is ultimately more than necessary in valid code stream. For some applications it may be undesirable. Two pass approach modifies the single-pass approach that supports only the values of significance and length, and discards any increment quality as the encoding of each image, and then, only as part of a valid code stream is determined (step 41), re-encodes all images (step 43)to generate when the stop quality which is used to generate valid code stream (step 42). During this second pass is possible to generate only the required increment of quality in order to avoid unnecessary calculations and/or unnecessary data storage.

An iterative approach

An iterative approach works as follows: for each image of each linked with the image of the subset are selected increment of quality, with the highest value of importance, so that the total size of the selected increment of the quality of the subset satisfy any associated with the image of the requirement for the size of this subset (step 34). Unselected increments quality are rejected. In addition, select all remaining increments of the quality of this image is owned by everyone associated with the sequence of the subset having values of significance above the threshold of significance for this subset (step 34). Unselected increments quality are rejected. For this last operation threshold value for each associated with a sequence of subsets supported fixed, until you have treated in this way all the images. The operation of step 34 may be performed simultaneously or can be ordered in a different way depending on the definitions under which nogusta. Once all images processed in this manner, the method determines whether the total size of the selected increment of quality associated with the sequence of the size requirements for associated with a sequence of subsets (step 37). If not satisfied, the process is repeated with changing thresholds of significance associated with a sequence of subsets (step 38) and returns to step 31, until you are satisfied associated with sequence size requirements, and then outputs a valid code stream (step 39).

Raising the threshold of significance is the lesser of the code flow, and lowering the threshold of significance is the larger. Often the requirements for different subsets are independent, and the search for suitable thresholds of significance can be performed independently. In this case, the threshold value can be changed at step 38 using a variety of methods, including the method of trial and error, divide in half, gradient descent or any other method of one-dimensional (1-D) numerical search. If the search cannot be carried out independently, can be used any way multidimensional numerical search. When indicators of inclination corresponding to the intensity of the distortion, are used as measures of significance, they can is to be used, to facilitate the search methods based on derivative/gradient.

The choice of increment quality

All options for the implementation of the above use the selection process increments quality, with the highest values of significance within the subset. This process can be performed in many ways. One way leads the list of increments quality in descending order of their values of importance and then selects from the top of the list. In another method, only the values of significance are listed in descending order as their compliance with the increments quality is known. In the next method, the threshold of significance established for a subset with the idea that you have selected all the increments quality with significance values above the threshold value. This may be seen as equivalent to the choice of the number of increments quality from the top of the ordered list of subsets when considering figure 3.

Figure 3 each solid horizontal line represents an increment of 50 quality, and each block of the horizontal lines represents the block 52 coding. A set of coding blocks is a subset of 54. The drawing size 56 significance is indicated next to each increment of quality. It is clear that the choice of all increments quality of each coding block, with the significance value above the threshold 58, fitted the frame in nineteen (as indicated by the underlined dashed horizontal lines), is equivalent to sorting all increments quality in the subset and then selecting fifteen increments from the top of the list. If the total length is chosen so increments quality is not required, the threshold value can be adjusted and the process is repeated until the total length is not required (within some tolerance). This is similar to an iterative process speed control described above, but in this case does not require the recording of data. Iteration occurs as part of step 34 or stage 41, using the previously calculated values of significance and length. This approach may have advantages in computation and data storage, in fact, sorting and leading the list of values the importance or increments quality. Raising the threshold of significance is the result of a smaller total size of the selected increment of quality, and lowering the threshold of significance is the result of a greater size. The threshold value can be changed in phase 34 and/or stage 41 with the use of a number of ways, including the method of trial and error, divide in half, gradient descent or any other method of one-dimensional (1-D) numerical search. Preferable to carry out the iteration is also possible to check out the many thresholds in parallel.

the parallel processing

As stated above, it is possible to perform the stages 31, 32 and 34 in parallel for multiple images. Also, it is possible to perform most of the work on the stages 41 and/or 42 in parallel. You can assume that each processor has an increment of quality, value relevance and dimensions for one or more images. Control processor can send to all processors in the magnitude of a threshold value for each subset. Each processor may return the total size for each subset of images in accordance with the transmitted values of thresholds. The control processor can summarize these dimensions and send new threshold values by iteration to achieve the desired overall size (total). When running then each processor can be requested to create and display a valid code stream for their images on the basis of increments quality, having the values of significance above the final threshold value.

High quality

Each approach strives to achieve high quality of the reconstructed image within the respective subsets (as doable within the requirements LOR). Specifically, embodiments of the stage 34 are making efforts to achieve high quality images. Options domestic is on the stage 36 are making efforts to achieve a high average quality for the full sequence. In each of the stages 34 and 36 embodiments of striving to achieve high quality through the inclusion of the increments quality, with the highest values of significance. In some cases, embodiments of reach approximately constant from image to image within the respective subsets. From the section above entitled "choice of the increment, it can be seen that single-pass, two-pass and iterative methods satisfy any associated with the sequence requirement for the specified related sequence of subsets through (somewhat equivalent) setting a threshold value for this subset and inclusion (of all images) all increments quality in the subset with significance values above this threshold.

Thus, valid code stream then contains all increments quality for all images in the framework associated with a sequence of subsets with significance values are above the threshold of significance is less than one, rejected at stage 34 by any appropriate associated with the image requirements. Thus, if no effect is associated with the image requirement, the quality will be (approximately) constant from image to image is ageney within the framework associated with a sequence of subsets. When we act and associated with the sequence, and associated with the image requirements, there can be significant changes in quality when a particular image is very difficult to encode than other in sequence. For this image increment quality are rejected in accordance with any related image requirement, and quality may be reduced for this image. The ability to achieve consistent quality, thus, is controlled according to the relative stringency of requirements related to the sequence against the image-related requirements. In particular, this ability decreases when associated with the image requirements are more stringent than the associated sequence requirements. In the extreme case, if no effect is associated with the consistency requirement can be considered that the desired sequence-related size was "as much as possible. In this case, associated with the image requirement is in effect for each image, and the quality will vary greatly.

Should also be noted that when the image is very simple to encode compared to other in sequence, as may be significantly higher than Takovo the other images, even though its compressed size can be very low.

Small size

As described above, in one-pass, two-pass and iterative versions of the implementation striving to achieve high quality permissible according to the size requirements in LOR. On the contrary, additional embodiments of striving to achieve a small size according to the admitted quality requirements (distortion) in the LOR. If for a subset of features associated with the image quality requirement, the stage 34 selects from a subset of the increment of quality, with the highest value of importance so that for a subset has satisfied any requirements for quality. If associated with a sequence of subsets of features associated with the sequence requirements for quality, one-pass and two-pass methods should be considered separately from the iterative method. For one-pass and two-pass methods stage 41 selects the increment of the quality of the associated sequence of a subset of all images so that the average quality of a subset (averaged over all images) to satisfy the requirements for quality. For an iterative method step 38 changes the threshold value, while the average quality will not meet the requirement for quality.

It is worth noting that when the project is the research Institute of the quality requirements, reduce the distortion associated with each increment of quality, can be useful in this regard (preferred lengths of the increments quality). It should also be noted that the types of requirements can be mixed. For example, the requirement for the size and quality can both be present in LOR for a single sequence.

Use layers to reduce the used RAM/external memory

Provided JPEG2000 mechanism layering can be included in a single pass or two pass of embodiments, in order to save RAM/external memory used to store values of significance and lengths, or to simplify some implementations. In this embodiment, the increment of quality are grouped to form two or more levels. Increment qualities that are more important, are placed on earlier layers, and higher quality, which are less important, are placed in the later layers. The layers can be formed so that each layer has the desired size. Alternatively, the layers can be formed so that each layer corresponds to a given threshold of significance, such as the permanent quality. In the embodiment with a single pass of the increment of quality also keep. In the embodiment, TLD is I passes increments quality are rejected. The layers then selects, for each subset to satisfy the requirements of the subset. Is it possible to merge layers or split the selected layers on new layers to achieve the desired structure tree view in valid code stream.

Layers to meet many LOR

In many cases, it will be possible to create a single code stream, containing many nested code threads are used, each for a different purpose. For example, a file with extremely high quality can be included in the archive version of the image sequence. From this file it may be possible to extract one or more different code streams, each designed for different applications. For example, one version may be intended for distribution with a high resolution movie, while another version may be intended for distribution with a lower resolution TV. This can be done without any "allocation rate" during checkout. The speed selection can be performed in advance. Removing just requires access to relevant data in the JPEG2000 format. As other examples for the different versions is possible to fit various sizes of compressed image, the foreground/back square is well etc., or their combinations.

This functionality can be achieved through the creation of a code stream with the presence of two or more layers. Defined subset, and set LOR, at least for one of the two or more layers. Outlined in the document how the selection of speed can be used to satisfy the LOR for each such layer. It is worth noting that you do not want the subsets were identical for each layer. Subset and LOR can set a different image for each layer according to the purpose of this layer.

In some cases, you may need to split the code stream with multiple layers into multiple files, each file contains data for one or more layers. For example, in the case of two layers, it may be useful to have data for the first and second layers in separate files. This will simplify the loading of only the first layer on the storage device (e.g., a server), with limited or performance when loading both layers on the storage device with a higher volume and/or performance. A similar discussion is supported for a variety of resolutions.

Examples of speed control

In the following examples using one-pass algorithms are illustrated different is combinatii associated with the image and associated with a sequence of subsets and requirements. In the examples described only size requirements, but the quality requirements can be satisfied using the above modifications.

In addition, the following examples for ease of discussion, refer only to the size of these increments quality. In practice, is required to take into account the size of the necessary service information, such as the basic headers, the headers pane/pane and the packet headers. This can usually be accomplished by a corresponding reduction of the budget of bits for data increment.

In conclusion, it should be noted that the examples do not address the detailed structure of the code JPEG2000 streams based on the selected increments quality. Selected increment qualities can be arranged in any manner permitted according to the JPEG2000 standard, including any order, sequence, etc.

Example 1: multi-component image

Consider a case where each image in the sequence has the color components "red", "green" and "blue".

Let,anddenotes the sets of block coding for components "red", "green" and "blue" imagerespectively. Let ,and.

Also letand.

Subsets for this example are

Associated with the image:,,,

Associated with the sequence:

All other sets above are intended only for convenience of notation.

The list of requirements is

Associated with the image:,,,

Associated with the sequence:(within tolerance)

whereis the maximum total size of any image,,andis the maximum total size of the component R, G and B of any image, andis the required General related to sequence the full size of the compressed code stream. In addition to the General requirement of size, individual images and/or components of each image should not require an excessive number of bytes. These requirements mouthbut useful to avoid overflow/buffer underrun or to limit the amount of computations performed in the decoder to decompress each image and/or components. The example is intended to cover the case when omitted one or more image-related subsets and their size requirements. The method also includes the case when you fall associated with a sequence of a subset and its requirement to size. Modifications to support these omissions included in the description below. The method is applicable to other color spaces such as XYZ. It is also applicable when using the color conversion. In this case, the requirements apply to (color) of the transformed components.

In a single-pass approach fromselected increment of quality, with the highest value of importance, so that. Unselected fromincrement quality are rejected (step 34). This process is also performed forand. Increment quality are selected from those remaining inthat have the most value, significance, so that the total size of the selected increments qualitywas less than or equal to(step 4). Unselected increments quality are rejected. Are any other order for the above samples.

When missing one or more of the requirements for,,andcorresponding part of the stage 34 is omitted. Once this process is completed for all images in the sequence, the increment of quality to choose fromwith the presence of the greatest values of significance towithin tolerance (step 41). If the associated sequence requirement is not present, the stage 41 can be omitted.

When there are associated with the sequence requirements, it is possible to replace(within tolerance). An implementation option above is modified only in thatmay be allowed to pass over severalat step 34. If you want an exact equality, this may be a requirement inequality, followed by filling insignificant bits. Unless there is associated with the sequence requirements and there are no requirements on, it is possible to install on, andthe demands of equality. Again an implementation option is almost unchanged. From this discussion, specialists in the art will be able to perform associated with the image requirements of equality and inequality for the other embodiments.

Example 2: multi-component image with variable resolution

Consider a case where each image in the sequence has the color components "red", "green" and "blue" and the code stream is generated to allow the display of decoded images at two different resolutions, for example 2KB and 4Kb. Approximate structure 130 code flow for this scenario is illustrated in figure 4 and the corresponding splitting 140 into frequency bands is illustrated in figure 5. Figure 4 code stream is segmented into six segments. Segments 1 and 4 contain compressed data from components "red", segments 2 and 5 contain the compressed data from the component "green", and segments 3 and 6 contain compressed data from components "blue". The first three segments allow the recovery image on the resolution of 2KB. The last three segments enable recovery image at a resolution of 4K, when used in conjunction with the first three segments. The structure of the code stream is intended for illustrating goals. The compressed data obtained on the basis of the algorithm can subsequently be made in the form of any valid code stream according to the JPEG2000 standard.

Let,anddenote the sets of block coding for components "red", "green" and "blue", respectively, which correspond to the frequency ranges 141, 142 and 143, respectively, contributing to the recovery image on the permissions and 2KB, and 4Kb. Let,anddenotes the sets of block coding for components "red", "green" and "blue", respectively, which correspond to the frequency ranges 144, 145 and 146, respectively, contributing to the recovery image only on the resolution of 4K. Let=,=and=denote the sets of all blocks coding, respectively, for component "red", "green" and "blue" in the image. Let= ,=and=denote the sets of all coding blocks that are in the image contribute and 2KB, and 4Kb, only 4Kb and full 4Kb, respectively. Let=,=and=denote the sets of all units coding for components "red", "green" and "blue", respectively, which contribute to resolution and 2KB, and 4Kb for the full sequence. Let=,=and=denote the sets of all units coding for component Kras is th", "green" and "blue", respectively, that contribute to the resolution of 4K only for the full sequence. Let=,=and=denote the sets of all units coding for components "red", "green" and "blue", respectively, which contribute to the full resolution 4K for the full sequence. Let=,=,=denote the sets of all coding blocks that contribute to and 2KB, and 4Kb, only 4Kb and full 4Kb, respectively, for the full sequence.

Subset for this example,

Associated with the image:,,,

Associated with the sequence:

All other sets above are intended only for convenience of notation

The list of requirements,/p>

Associated with the image:

,,S (the total size of R, G, B 2KB)

(the total size of the full 4Kb)

Associated with the sequence:

(the total size of the full 4Kb) within tolerance

In addition to the General information associated with the sequence of the requirement for full size 4Kb separate image, and the portion in 2KB individual color component should not require excessive amounts of bytes. These requirements can be useful to avoid overflow/buffer underrun or limit the amount of calculations performed for decompression decoder. The method is applicable to other color spaces such as XYZ. It is also applicable when using the color conversion. In this case, the requirements apply to (color) of the transformed components. Finally, the method is applicable when there is one or more image-related requirements. In this case, can be passed through the appropriate parts of the stage 34. In this way the algorithm is also applicable, when is not associated with the sequence requirement. In this case, can be passed through the appropriate parts of the stages 34, 41, 37 and/or 38.

In a single-pass approach from select increments quality, with the highest values of significance, so. All unselectedincrement quality reject. The same operation is done forand. Then, the increment of quality chosen from those remaining inhaving the greatest values of significance, so that. Unselected increments quality reject.

After all images are processed in such a way, fromselect increments quality, with the highest values of significance, sowithin tolerance (step 41). Unselected increments quality reject. The selected increment of quality then used to generate a code stream. If not there is associated with the sequence requirement, the stage 41 can be omitted.

It is of interest to consider the formation of a code stream from this example in two layers. One way of satisfying the above requirements for these two layers together by selecting increments as described above. These increments quality are divided into two layers. A simple case may requireandfor the first layer, moreover, the <and<. Other cases are similar. For this case, the increment of quality for the first layer can be selected as set forth below. For each image increment quality, selected as described above and having the greatest values of significance is chosen so that. Then, among all selected so choose increments quality so that. This last selection process so thatyou can skip if on the first level there is associated with the sequence requirements. Increment quality, selected, fill the first level, and the remainder goes to the second level. Specialists in the art will be able without difficulty to extend this implementation to include many layers.

Another approach to the inclusion of two layers is that the above situation is reversed. These two layers together satisfyand (possibly)and>and>. Then select the first layer, in order to meet the requirements as original is set forth in example 6. Forward-only way to meet these requirements is set out below. First, choose the increment of quality in order to satisfy the requirements for these two layers together. Specifically, for each image increment of quality, with the highest value of importance, are chosen so that, discarding all the rest.

How, then, of all remainingincrements quality picks with the highest value of importance in order. (The choice to satisfyyou can skip, if this requirement is not present.) All increments quality, selected, enter a valid code stream. The remainder is discarded. Increment of quality in valid code stream, which belong to the first layer, can be selected using the above variants of implementation example 2 to meet initial requirements. Expanding on the number of layers more than two immediately.

Example 3: Alternative values of significance

A common way to perform speed control schemes unscaled compression is with the initial quantization step of the quantization and encoding data of all coefficients, then the iteration quantization and encoding for others the number of quantization steps, until requirements are met. This is usually used for only one subset (containing the full image). "Single-pass" option implementation of the present invention can achieve the same effect for many subsets and/or many of the requirements through the use of quantities of bit planes as the value of significance. The allocation of the resulting velocity makes an effort to maximize the number of bit planes that are included from the respective subsets, taking into account LOR. The result will be equivalent to using a fixed quantization step of the quantizer for each coefficient in the subset. The effective size of the quantization step is then 2pΔ, where Δ is the actual size of the quantization step used in the step 31, and p is the number of bit planes, rejected within the subset. This approach may require a large tolerance on the required associated with a sequence of dimensions. If this is unacceptable, then you can iterate. Probably that is required only one additional pass. This is because it is guaranteed that the desired final value of Δ will lie between the 2pΔ 2p-1Δ, where p is the minimum number of bit planes is if rejected (step 41), give associated with the sequence size is less than the required. You can then use interpolation of these two values so that the following passage was probably issued associated with the sequence size within appropriate for the desired size of the tolerance. The interpolation can be linear or nonlinear. In this regard, it may be useful dependence that the value Δ is often proportional to Csizefor some constant C.

Other examples

Described in detail above examples represent only some of the possible design solutions for speed control and some of the possible requirements. For example, each image in the sequence can be spatially segmented into two regions, called foreground and background, which is not required to be connected, and is not required to be static between images, and which have different levels of quality after decompression. In this case, the coding blocks corresponding to the foreground and the background, divided into different subsets and treated with single-pass, two-pass or iterative algorithm. One possible requirement may be that the total size of the compressed data introduced by the coding blocks in the subset j is between the very upper limit value and lower limit value. Also, the requirements may vary for groups (mini-sequences) images. This can be useful for applications editing or transmission over time-varying communication channels.

Illustrative of the application of speed control

There are many other possible applications are described in the document methods of speed control to generate code streams, which represent high average quality of the reconstructed image or the small size of the compressed when addressing the LOR for subsets of images. Below are some illustrative applications.

Speed control for modifications after the release of

Suppose, after coding sequence is required to replace one or more images. Each image replacement can be separately encoded to have the same number of bytes in each subset, such as for the image, which it replaces, using the guidance of the present invention. Alternatively, image replacement can be encoded as a sequence, even if they may be from different time locations within the total sequence. Associated with the sequence required dimensions for "sequence" replacement can be chosen to correspond to those for C the change of images.

Speed control for threading backup and/or compressed image

In the process of simultaneous consideration of the patent application U.S. serial No. 11/051771, filed February 4, 2005, hereinafter referred to as "DI Workflow" (continuous processing of digital images), which thereby is fully incorporated herein by reference, describes a continuous processing of the compressed JPEG2000. Described in the document methods of speed control can be used in this environment. Any editing of the sequence of images and/or images in the sequence can be performed using the uncompressed data before any allocation of the coding rate, and transmission. But preferably, as shown in Fig.6 and 7, JPEG2000 encoder 150, or more generally scalable encoder encodes each image 152 (step 154) (preferred for very high quality, is possible without losses) and stores the values of significance and length (step 158) for all increments quality in accordance with the stages 31 and 32 of figure 2. Increment the quality of compressed images also stores (preferably in the form of code JPEG2000 streams, but are also available in other data structures) (step 160).

The stored data can be used as a high-quality compressed archive for the latter the total allocation speed and/or editing. If you want to edit, images can be decompression, edited, and then re-compressed, or editing operation (step 162) can be performed by the workstation 164 DI using compressed code streams, as described in the DI Workflow. The values of significance and retain length (step 166) for any newly formed (replacement) increment of quality. Once the editing is completed, the stored data can again be used as a high-quality archive. When you want to perform the selection of the speed controller 168 speed applies methods (step 170) speed control to the edited (or stored) of the code stream with the given definitions 172 and LOR 174 for subsets and generate a valid code stream 176 (step 178), which can be archived or recorded on a storage medium or transmitted over the channel (connection).

To perform speed control is possible to decompress and then recompress using methods as described in accordance with the present invention. Preferably, does not perform decompression/re-compression. Single-pass algorithm may simply act, performing the step 34 in respect of all compressed images, then go to step 36. Alternatively, since all shows what I already compressed and stored, stages 34 and 36 can be run simultaneously. It is clear that thus cannot be created more than one valid code stream for each version corresponding to a different LOR. Can be supported entry for each code stream generated in this way, without saving the generated code themselves flows. One such method stores the corresponding thresholds of significance used to meet every requirement associated with the image and associated with the sequence. As shown in the configuration of figure 6 and 7, the operation of the compression and editing and speed control are performed by JPEG2000-encoder 150, workstation 164 DI and controller 168 speed. However, these operations can be integrated into a single workstation.

An implementation option above describes the conservation of the increments, together with their significance values and lengths. These lengths are useful in meeting the size requirements. If instead (or in addition) subject to the satisfaction of quality requirements, can be saved values reduce distortion to increment quality instead of (or in addition to) length increment.

Additional speed control to create archive

As described above, the archive can contain all of the increment of the quality with their values testing the cost and lengths. This provides greater flexibility to create later a different valid code streams for the other subsets and definitions LOR (possibly unknown during the execution of the stages 31 and 32). On the other hand, as soon as all interest subsets and determine LOR are known, you can create and store a valid code flow is very high quality, it is possible without loss, as an additional archive. When use is described in the document (using one simple case shown in example 2) methods of speed control can be created archive containing various layers that meet different LOR for different definitions of the subsets. Different valid code streams that satisfy different LOR, then can be removed in the future without the need for any additional allocation speed.

Then as been shown and described several illustrative embodiments of the invention, numerous variations and alternative embodiments of will be clear to experts in the given field of technology. For example, it is possible to apply the embodiments of this invention to a subset of the images in the sequence to reduce computation. The initial set of parameters calculated using the subset of all images then can be applied to the full image sequence. This technique can result in significant savings calculations. In addition, some embodiments of can be applied to a single image. Additionally, options for implementation are applicable when using temporal prediction or a temporary transformation. In these cases, the methods are applied to images with forecast error or converted in time images. In the case of temporary conversion of a subset can be defined in relation to groups converted in time images. For example, consider a three-level temporal wavelet transform. In this case, eight converted in time images correspond nominally eight original images. Thus, it may be useful to identify subsets that are applicable to the eight converted in time images. Such variations and additional embodiments of assumed, and they can be made without departing from the scope of the essence and scope of the invention defined in the attached claims.

1. The method of speed control for one or more images, converted to obtain the conversion coefficient, which is divided into blocks to which tiravanija, each of which is encoded in multiple increments quality, having respective values of significance, containing phases in which:
define subsets as sets of block coding, comprising the first subset and the second subset, which include a different set of coding blocks, and at least one of the first and second subsets includes two or more blocks of the coding, and at least one image delivers one or more blocks of the coding in each of the first and second subsets;
establish a list of requirements (LOR) for these subsets, and LOR this includes the first requirement to the size of the first subset and the second requirement to the size of the second subset, while the first and second size requirements differ;
use the values of significance for choice with the help of the controller speed increments of a quality to satisfy the first requirement for the size of the first subset and the second requirement for the size of the second subset; and
create valid code stream for the said one or more images based on the selected increments.

2. The method according to claim 1, in which at least first and second subsets include the same coding block.

3. Pic is b according to claim 1 or 2, in which images are encoded using JPEG 2000, where the units of encoding are subject to encoding blocks of coefficients and increments are coded passages.

4. The method according to claim 1 or 2, in which a valid code stream includes two or more layers, and one or more layers together satisfy the first and second size requirements.

5. The method according to claim 1 or 2, in which each increment of quality has length, choose the increment of quality with the highest value of importance as long as the total length of the selected increments quality will not satisfy the requirement for the size of the subset.

6. The method according to claim 1 or 2, in which at least one element in LOR is required to quality for a subset, thus choose the increment of quality with the highest values of significance with regard to the quality requirements.

7. The method according to claim 1, wherein the first subset represents is associated with the image of a subset, defined as a set of blocks of the encoding based on the individual image and the second subset is associated with a sequence subset, defined as sets of blocks coding on the basis of each image in the sequence of many of the above images, and LOR who engages in itself associated with the image requirement and/or associated with the sequence requirement, these increments quality is chosen according to: for the first subset and any other image-related subsets are selected via the speed controller increments quality to meet them associated with the image requirements of the LOR; and for the second subset, and any other related sequence subsets are selected via the speed controller increments quality to satisfy their related sequence requirements of the LOR.

8. The method according to claim 1, wherein the first and second subsets are associated with a sequence of subsets, defined as different sets of coding blocks from each image in the sequence of many of the above images.

9. The method according to claim 1, wherein the first and second subsets are associated with the image subsets, defined as different sets of block coding based on individual image.

10. The method according to claim 7, in which the total valid code stream for these multiple images created in one pass.

11. The method according to claim 10, in which the increment of the quality chosen by using the speed controller on the basis of their significance values to suit associated with the image request for the first subset and the selected priase the Oia retain quality for each of the above mentioned image, and then choose by using a speed controller increments quality based on values of significance to suit associated with the sequence requirement for the second subset, and then create valid code stream based on the selected stored increments.

12. The method according to claim 7, in which associated with the image size requirements for the first and second subsets associated with the image, specify the maximum size or the desired size.

13. The method according to claim 7, in which the said subsets include for each image first, second and third associated with the image of a subset, which respectively include all blocks of the coding necessary to restore each of the three color component image, and the fourth is associated with the image of a subset, which is the Union of three color subsets, while LOR includes associated with the image requirements to the maximum size for each of the first, second and third image-related subsets, which specify lower maximum allowable size than allowed according to the associated image to the size for the fourth associated with the image subset.

14. The method according to item 13, in which the said subset further what about the include associated with the sequence of the subset, includes all the coding blocks all images of the sequence, and LOR includes associated with the sequence requirement for the size of these is associated with a sequence of subsets, which allows larger than associated with the image requirement to maximum size for the fourth associated with the image subset.

15. The method according to claim 7, in which a valid code stream provides an opportunity to restore the image on the first lower resolution and at a second higher resolution, these subsets include for each image first, second and third associated with the image subsets that include all the blocks of the coding necessary to restore each of the three color components of an image on the first lower resolution, and the fourth is associated with the image of the subset that includes all of the blocks encoding the full image, while LOR includes for each image associated with the image size requirement for the fourth associated with the image subset and associated with the image requirements to the maximum size for each subset of the first, second and third image-related subsets that require Myung-the-big-sizes than allowed pursuant to the requirement of the size of the fourth subset.

16. The method according to item 15, in which a valid code sequence includes two or more layers, and one or more layers together satisfy the above requirements.

17. The method according to item 15, in which the mentioned associated with the image of the requirement for the size for the fourth associated with the image subset is associated with the image requirement to maximum size.

18. The method according to 17, in which the above-mentioned subset additionally include associated with the sequence of the subset that includes all of the coding blocks all images, and LOR includes associated with the sequence requirement for the size of these is associated with a sequence of subsets.

19. The method according to claim 7, in which for each image the first and second subsets are associated with the image subsets, which contain blocks of coding required to restore the different spatial regions of the image.

20. The method according to claim 1 or 2, further containing a sequence that includes several images, the steps that remain in the archive referred to increment quality blocks for encoding, thus referred to increment the quality of choose from this archive, to meet the LOR for each subset and create valid code stream.

21. The method according to claim 20, in which the associated values of significance remain in the archive.

22. The method according to claim 20, in which the steps of identifying subsets and installation LOR perform after the increment of the quality of the saved in an archive.

23. The method according to claim 20, further comprising stages, which are:
set the other LOR for the same or other subsets for the same image sequence; and
use the values of significance to choose the increment of the quality of the archive to create another valid code stream that meets this other LOR.

24. The system of formation of a valid code stream containing:
the encoder is configured to accept an input sequence of one or more images and convert them in order to obtain transform coefficients, which are divided into coding blocks, each of which is encoded in multiple increments quality, having respective values of significance and length; and
the speed controller is made with the ability to define subsets of the sets of blocks of encoding that includes the first subset and the second subset comprising different sets of the locks encoding, at least one of the first and second subsets includes two or more blocks of the coding, and at least one image delivers one or more blocks of the coding in each of the first and second subsets, and make a list of requirements (LOR) for these subsets, which includes the first requirement to the size of the first subset is different from the second size requirements for the second subset, and the speed controller uses the values of significance to choose the increment of quality that meet LOR for the relevant subsets, and to create valid code stream based on the selected increments quality.

25. The system of paragraph 24, in which at least first and second subsets include the same coding block.

26. The system of paragraph 24 or 25, in which the said subsets include for each image first, second and third associated with the image of a subset, which respectively include all blocks of the coding necessary to restore each of the three color component image, and the fourth is associated with the image of a subset, which is the Union of three color subsets, while LOR includes associated with the image requirements to the maximum size for the each of the first, the second and third image-related subsets, which specify lower maximum allowable size than allowed according to the associated image to the size for the fourth associated with the image subset.

27. The system of paragraph 24 or 25, in which the valid code stream provides an opportunity to restore the image on the first lower resolution and a second higher resolution, and the said subsets include for each image first, second and third associated with the image subsets that include all the blocks of the coding necessary to restore each of the three color components of an image on the first lower resolution, and the fourth is associated with the image of the subset that includes all of the blocks encoding the full image, while LOR includes for each image associated with the image size requirement for the fourth associated with the image subset and associated with the image requirement to the maximum size for each of the first, second and third image-related subsets that require smaller size than allowed pursuant to the requirement of the size of the fourth subset.



 

Same patents:

FIELD: information technology.

SUBSTANCE: coding device has definition apparatus for determining image area data meant for processing in order to counter reconstruction implied by granular noise arising in image data coded based on said image data and apparatus for countering reconstruction, designed for processing in order to counter reconstruction for image area data, defined using definition apparatus when coding image data, where when the said image data are coded in data unit data modules, the said definition apparatus determines unit data which form the said image data as the said image area data, and apparatus for countering reconstruction forcibly sets the orthogonal transformation coefficient to zero, which becomes equal to zero when quantisation is carried out using the said unit data, among orthogonal transformation coefficients of unit data defined using the said definition apparatus.

EFFECT: improved quality of the decoded image.

14 cl, 18 dwg

FIELD: information technologies.

SUBSTANCE: method includes the following: operation of coding for formation of coded images in coder, and also operation of specified coded images transfer in decoder in the form of transmission units, operation of buffering for buffering of transmission units sent to decoder, in buffer and operation of decoding for decoding of coded images with production of decoded images. Buffer size is specified by means of determining overall size of at least two transmission units and setting maximum size of buffer on the basis of this overall size.

EFFECT: improved efficiency of coded images buffering.

22 cl, 16 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: physics; image processing.

SUBSTANCE: invention relates to the technology of simulating granularity of a film in an image. A method is proposed for simulating a film granularity unit for adding to an image unit through a first establishment of at least one image parametre in accordance with at least one unit attribute. The film granularity unit is established in accordance with the image parametre. Deblocking filtration can also be applied to the film granularity unit.

EFFECT: easier film granularity simulation.

27 cl, 4 dwg

FIELD: physics; communications.

SUBSTANCE: invention relates to an encoder/decoder of a scaled data stream, which includes at least two scalability levels. Proposed are a method and a device for encoding, decoding, storing and transmitting a scaled data stream, which includes levels with various encoding characteristics. The method involves creating one or more levels of a scaled data stream. At least one level has an encoding characteristic which includes at least one of the following: information on fine granular scalability (FGS); information on region of interest (ROI); information the scaled sub-sample level; information on relationships of decoding and the set of initial parameters. The method also involves signalling levels using encoding characteristics such that, the characteristics can be read by the decoder without need for decoding an entire level.

EFFECT: increased efficiency of a scaled data stream and possibility of direct transmission in a bit stream in a file format or through a transfer protocol of information scalability for a specific level of a scaled bit stream.

49 cl, 5 dwg, 8 tbl

FIELD: physics; image processing.

SUBSTANCE: invention relates to encoding and decoding video data, and more specifically to scaled video processing. A method is proposed for scaled video data encoding, which involves receiving said video data, after which, based on the received video data, a basic level is generated, which includes at least one image and at least one enhancement level which includes at least one image. For each of the said basic levels and enhancement level, a characteristic identifier is generated, which is associated with a reference number; the corresponding sequence parametre set (SPS) is determined for each of the basic levels and at least one enhancement level, with different values of characteristic identifier. For a basic level and enhancement level with same SPS parametres, one sequence parametre set is used; and said basic level and said at least one enhancement level are encoded by using sequence parametre sets which determined.

EFFECT: increased encoding or decoding efficiency, avoiding redundancy.

12 cl, 10 dwg

FIELD: physics; image processing.

SUBSTANCE: invention relates to methods of imitating film granularity in an image. The said result is achieved due to that, film granularity in a video image is imitated by creating a block first i.e. a matrix of transformed coefficients for a set of cutoff frequencies fHL fVL, fHH and fVH, related to the desired granularity structure. Cutoff frequencies fHL fVL, fHH and fVH represent cutoff frequency, in two measurements, of a filter which sets characteristics of the desired film granularity structure. The block of transformed coefficients undergoes inverse transformation, obtaining a sample of film granularity with accuracy of up to a bit, and the said sample is scaled in order to mix with a video signal for imitating film granularity in that signal.

EFFECT: easier imitation of film granularity in an image.

20 cl, 4 dwg

FIELD: information technologies.

SUBSTANCE: imitation of film grain block to be added to picture block is first performed by setting at least one parametre at least partially in compliance with picture block attribute. At least one film grain block is formed from at least one film grain structure, which is generated in compliance with at least one parametre. Film grain structure is namely generated by using the method accurate to a bit.

EFFECT: decreasing film grain imitation complexity in the picture in order to be reproduced in media reproducing devices.

34 cl, 4 dwg

FIELD: information technologies.

SUBSTANCE: film grain structure accurate to a bit is designed by means of organising, first, the set of converted coefficients accurate to a bit. Set of converted coefficients accurate to a bit is subject to frequency filtering and the following reverse conversion accurate to a bit in order to finally obtain film grain structure.

EFFECT: reducing film grain imitation complexity in the picture.

17 cl, 6 dwg

FIELD: information technologies.

SUBSTANCE: there proposed is method and device for reading motion picture film grain structures in raster sequence when imitating the grain of motion picture film including the setting of pseudo-random initial position, repetition of pseudo-random initial position for each line of the motion picture film grain block group and use of the other pseudo-random initial position for each line of presentation of the following group of motion picture film grain blocks. In various design versions of this invention, other pseudo-random initial positions are initiated by means of repeated setting of at least one reference value of pseudo-random number generator made for determining the above pseudo-random initial positions.

EFFECT: reducing the film grain imitation complexity in the picture and providing the possibility of interpreting the information message of additional improvement of motion picture film grain.

24 cl, 6 dwg

FIELD: information technology.

SUBSTANCE: coding device has definition apparatus for determining image area data meant for processing in order to counter reconstruction implied by granular noise arising in image data coded based on said image data and apparatus for countering reconstruction, designed for processing in order to counter reconstruction for image area data, defined using definition apparatus when coding image data, where when the said image data are coded in data unit data modules, the said definition apparatus determines unit data which form the said image data as the said image area data, and apparatus for countering reconstruction forcibly sets the orthogonal transformation coefficient to zero, which becomes equal to zero when quantisation is carried out using the said unit data, among orthogonal transformation coefficients of unit data defined using the said definition apparatus.

EFFECT: improved quality of the decoded image.

14 cl, 18 dwg

FIELD: physics; image processing.

SUBSTANCE: invention relates to an image processing device for converting data of a moving image having a first frame frequency, into data of a moving image having a higher frame frequency. The technical result is achieved due to that, a low-pass filter (LPF) filters image input data frames (A[i]) so as generate low-frequency image data. A subtracting unit and an adder generate high-frequency image data. Another adder adds low-frequency image data coming from a delay circuit, with subsequent low-frequency image data. A divider divides the sum by two to generate low-frequency averaged image data. A switch successively outputs high- and low-frequency image data each time an image data frame is entered.

EFFECT: device can generate output data image with double the frequency of input image data frames; easier image processing and reduced blurring in a retaining type display device during movement and reduced flickering in a pulse type display device.

17 cl, 32 dwg

FIELD: physics; image processing.

SUBSTANCE: invention relates to the technology of simulating granularity of a film in an image. A method is proposed for simulating a film granularity unit for adding to an image unit through a first establishment of at least one image parametre in accordance with at least one unit attribute. The film granularity unit is established in accordance with the image parametre. Deblocking filtration can also be applied to the film granularity unit.

EFFECT: easier film granularity simulation.

27 cl, 4 dwg

FIELD: physics; image processing.

SUBSTANCE: invention relates to devices for simulating film grains. In order to determine film grain parametres, apparatus for determining film grain receives an input information stream and a filtered information stream from which film grain has been removed. The apparatus for determining film grain parametres gives out a message from these streams, which contains a model identifier for grain simulating, as well as at least one set from several parametres which include correlation parametres, intensity-independant parametres and intensity-dependant parametres used by the identified model. An encoder encodes film grain information for subsequent transmission.

EFFECT: simulation of film grain for mixing with an image after decoding.

26 cl, 3 dwg

FIELD: physics; image processing.

SUBSTANCE: invention relates to methods of imitating film granularity in an image. The said result is achieved due to that, film granularity in a video image is imitated by creating a block first i.e. a matrix of transformed coefficients for a set of cutoff frequencies fHL fVL, fHH and fVH, related to the desired granularity structure. Cutoff frequencies fHL fVL, fHH and fVH represent cutoff frequency, in two measurements, of a filter which sets characteristics of the desired film granularity structure. The block of transformed coefficients undergoes inverse transformation, obtaining a sample of film granularity with accuracy of up to a bit, and the said sample is scaled in order to mix with a video signal for imitating film granularity in that signal.

EFFECT: easier imitation of film granularity in an image.

20 cl, 4 dwg

FIELD: physics; image processing.

SUBSTANCE: invention relates to digital video information processing, and more specifically to methods of coding and decoding images, and is meant for designing systems for coding and decoding based on three-dimensional discrete cosine transformation of video data. To detect and eliminate time redundancy in each domain with size n×n×n pixels, discrete cosine transformation with respect to time is carried out. Presence of movement in each fragment with size n×n pixels is then determined from presence of non-zero spectrum factors except the first fragment of the domain. If there is movement in each fragment of the domain, in order to eliminate spatial redundancy, coefficients of discrete cosine transformation are calculated on two spatial coordinates x and y. The obtained coefficients are quantized and coded with elimination of statistical redundancy and then transmitted to a communication channel. During decoding, the entire process is carried out in reverse order.

EFFECT: increased efficiency of processing video information and increased compression ratio of video data using 3D discrete cosine transformation over a time interval t and formation of a buffer for storing information on domains without movement.

4 dwg

FIELD: information technologies.

SUBSTANCE: imitation of film grain block to be added to picture block is first performed by setting at least one parametre at least partially in compliance with picture block attribute. At least one film grain block is formed from at least one film grain structure, which is generated in compliance with at least one parametre. Film grain structure is namely generated by using the method accurate to a bit.

EFFECT: decreasing film grain imitation complexity in the picture in order to be reproduced in media reproducing devices.

34 cl, 4 dwg

FIELD: information technologies.

SUBSTANCE: film grain structure accurate to a bit is designed by means of organising, first, the set of converted coefficients accurate to a bit. Set of converted coefficients accurate to a bit is subject to frequency filtering and the following reverse conversion accurate to a bit in order to finally obtain film grain structure.

EFFECT: reducing film grain imitation complexity in the picture.

17 cl, 6 dwg

FIELD: information technologies.

SUBSTANCE: there proposed is method and device for reading motion picture film grain structures in raster sequence when imitating the grain of motion picture film including the setting of pseudo-random initial position, repetition of pseudo-random initial position for each line of the motion picture film grain block group and use of the other pseudo-random initial position for each line of presentation of the following group of motion picture film grain blocks. In various design versions of this invention, other pseudo-random initial positions are initiated by means of repeated setting of at least one reference value of pseudo-random number generator made for determining the above pseudo-random initial positions.

EFFECT: reducing the film grain imitation complexity in the picture and providing the possibility of interpreting the information message of additional improvement of motion picture film grain.

24 cl, 6 dwg

FIELD: physics.

SUBSTANCE: imitation of film grain inside a receiver takes place by obtaining at least one unit of pre-calculated modified coefficients first. The said unit of modified coefficients undergoes filtration, corresponding to a frequency range which characterises the desired film grain structure. In practice the frequency range lies within the limits of the set of cut-off frequencies fHL, fVL, fHH and fVH of the filter in two measurements, which characterises the desired film grain structure. After this the filtered set of coefficients undergoes inverse transformation to obtain the film grain structure.

EFFECT: easier imitation of film grain in an image.

15 cl, 4 dwg

FIELD: technology for processing remote probing data for detection and recognition of objects on basis of their images.

SUBSTANCE: method includes source geometrically pixel-wise combined digitized images of one and the same scene concurrently in n spectrum ranges, source signs matrix is formed, each element of which represents n-dimensional values vector of signals of pixels of source images with similar coordinates, standard is selected in form of arbitrary element of source signs matrix, final numeric matrix is formed, to each current element of which value is assigned, equal to distance in vector signs space between vector, appropriate for standard and vector, appropriate for element of source matrix with same number of row and column to those of current element, final matrix is transformed to digital image, while as signs, brightness values of pixels are used, textural and gradient characteristics of source image pixels.

EFFECT: simplified operations concerning generation of synthesized image for visual interpretation, its adaptation to objects targeted by observer, detailed reflection of chosen objects on synthesized image and compact representation of information.

4 cl, 2 dwg

Up!