256 results about "Nonzero coefficients" patented technology

A block transform-based digital media codec more efficiently encodes transform coefficients by jointly-coding non-zero coefficients along with succeeding runs of zero-value coefficients. When a non-zero coefficient is the last in its block, a last indicator is substituted for the run value in the symbol for that coefficient. Initial non-zero coefficients are indicated in a special symbol which jointly-codes the non-zero coefficient along with initial and subsequent runs of zeroes. The codec allows for multiple coding contexts by recognizing breaks in runs of non-zero coefficients and coding non-zero coefficients on either side of such a break separately. The codec also reduces code table size by indicating in each symbol whether a non-zero coefficient has absolute value greater than 1 and whether runs of zeros have positive value, and separately encodes the level of the coefficients and the length of the runs outside of the symbols.

An improved video compression system in which the coefficient transform is sped up via folding transposition of coefficients into the token extraction phase. This is accomplished by filling a group of blocks coefficient buffers with 0's before the start of coefficient decoding, extracting the token from the bitstream and placing any extracted coefficient value which is non zero into the transposed order that best suits the target processor.

An image coding method of compressing and coding image data includes: binariziang a plurality of coefficients to generate a binary signal, the plurality of coefficients being included in a unit of processing of the image data in a frequency domain; determining a context to be used for arithmetic coding of the plurality of coefficients, based on a position of a last non-zero coefficient in a scan order among one or more non-zero coefficients included in the unit of processing; performing arithmetic coding on the binary signal using probability information corresponding to the determined context; and updating, based on the binary signal, the probability information corresponding to the determined context.

A method, program product and apparatus for encoding a scalable bit stream from the binarization results of a video sequence by selectively encoding syntax elements and avoiding redundancy in coding. The result is a decrease in the size of the compressed bit stream of an enhancement layer. One method includes determining whether a skipping flag in the base layer macro block of the video data is set, and encoding an enhancement layer macro block of the video data, corresponding to the base layer macro block, with a skipping flag only if the base layer macro block skipping flag is set. Another method includes determining which of a plurality of blocks in a base layer macro block contain zero coefficients, generating a coded block pattern (CBP) of an enhancement layer macro block, where the CBP includes a number of digits equal to the number of blocks in said base layer macro block containing only zero coefficients, and then encoding the CBP of the enhancement layer. Yet another method includes encoding a CBP value of a base layer macro block and differentially encoding a CBP value of an enhancement layer macro block relative to the CBP of the base layer macro block. An additional method includes determining the zero-value coefficients in a block of a base layer, determining whether any of the zero-coefficients become non-zero coefficients in a corresponding block in an enhancement layer, and encoding a coding block flag in an enhancement layer based on that determination.

Transform coefficients of sample blocks of a macroblock of a video picture are encoded by adaptively encoding in combination, the number of non-zero coefficients before the trailing one coefficients and the number of trailing one coefficients. The transform coefficients may be further encoded by adaptively encoding one or more of the signs of the trailing one coefficients, the level measures of the non-zero coefficients, the total number of zero coefficients interposed in the non-zero coefficients, and the respective run lengths of the zero coefficients. Adaptive encoding of the number and trailing one coefficients may be performed in view of one or more neighboring sample blocks, whereas adaptive encoding of level measures may be performed in view of quantization parameters of a macroblock and previously encoded level measures. Decoding may be performed in an inverse manner.

Entropy encoding is performed in the inventive apparatus and method in response to the scanning of transform coefficients following an initial scanning pattern selected on the basis of probability statistics of non-zero coefficients for each block position. These non-zero probability statistics are ranked for a given combination of coding characteristics within the current block to arrive at an initial scanning pattern. The same initial scanning pattern selection is performed in the decoder to allow the transform coefficients to be extracted in their proper order from encoded video data. The pattern selection is applicable to both intra prediction and inter prediction. Transform coefficients are more accurately ordered in response to the invention because in adapting pattern initialization to quantization step size, high-frequency basis functions are properly taken into account.

The method includes following steps: scanning DCT coefficient blocks according to sequence so as to obtain last one nonzero coefficient (NC) and the coefficient in front of the last one NC in coefficient block, and number value of NC; based on the number value of NC to calculate difference value between number value of NC in this block and number value of NC in adjacent block, and constructing context model for number value of NC; calculating flag bit of NC for last one NC and coefficients before the last one NC; selecting context model to encode flag bit of NC; subtracting 1 from absolute value of NC according to scanning sequence or scanning inverted sequence; and selecting context model to encode number value of NC, and its position in the scan etc; encoding sign bit. The invention also discloses corresponding decoding method.

A method, program product and apparatus for decoding from a scalable bit stream the binarization results of a video sequence by selectively decoding syntax elements and avoiding redundancy in coding. The result is a decrease in the size of the compressed bit stream of an enhancement layer bit stream. One method includes determining whether a skipping flag in the base layer macro block of the video data is set, and decoding a skipping flag from an enhancement layer macro block of the video data, corresponding to the base layer macro block, only if the base layer macro block skipping flag is set. Another method includes determining which of a plurality of blocks in a base layer macro block contain zero coefficients, decoding an abbreviated coded block pattern (CBP) of an enhancement layer macro block, where the CBP includes a number of digits equal to the number of blocks in said base layer macro block containing only zero coefficients, and then generating a complete CBP for the enhancement layer based on the results of said decoding. Yet another method includes decoding a CBP value of a base layer macro block and differentially decoding a CBP value for an enhancement layer macro block relative to the CBP of the base layer macro block. An additional method includes determining the zero-value coefficients in a block of a base layer, decoding a coded block flag for a corresponding block in an enhancement layer, and determining, based on the value of said encoded block flag, whether any of the zero-coefficients become non-zero coefficients in said enhancement block.

Systems and methods for quantization are provided. Some embodiments provide a system and method for quantization comprising preprocessing the transform coefficients to predict one or more non-zero coefficients and one or more zero coefficients as well as predict the non-existence of non-zero and zero coefficients, storing indices representing the predicted non-zero coefficients, and performing a quantization process on the predicted non-zero coefficients, as well as the inverse quantization process of those non-zero quantized coefficients.

In wavelet-based image compression schemes, a very sparse representation of the image signal may be obtained after quantization to transform coefficients. In addition, the nonzero coefficients 2-dimensionally cluster around the edge or texture areas. In existing systems, for example JPEG2000, in the bit-plane coding process, coefficients are repeatedly scanned and encoded in a 1-dimensional pattern within code-blocks. A large number of zeros have to be encoded to record the distribution of significant coefficients. It inevitably causes a big loss of compression performance. Quaternary reaching method emphasizes reaching and encoding the significant coefficients in 2-dimensional pattern. It fully adapts to the 2-dimensional character of the significance distribution of quantized coefficients. Besides, it admits very economical implementation. The recording of redundant information is drastically reduced. As result, it magnificently enhances both compression performance and computation performance against existing systems.

An optimal scanning method for coding / decoding an image signal is provided. In a method of coding an image signal through a discrete cosine transform, at least one is selected among a plurality of reference blocks. A scanning order in which to scan blocks to be coded of the reference blocks is generated and the blocks to be coded are scanned in the order of the generated scanning order. The at least one selected reference block is temporally or spatially adjacent to the block to be coded. When the blocks to be coded are scanned, probabilities that non-zero coefficients occur are obtained from the at least one selected reference block, and the scanning order is determined in descending order starting from the highest probability. Here, the scanning order is generated to be a zigzag scanning order if the probabilities are identical. The optimal scanning method increases signal compression efficiency.

The invention belongs to the field of picture / video coding / decoding in the signal processing, relating to an entropy coding method of conversion coefficient in the picture / video coding. The invention extracts the position information and value information of non-zero coefficients from a string of coefficients scanned to make the variable-length coding, and as coding the position information, combines the continuous zero coefficients and non-zero coefficients into two- dimensional combined pairs of zero-coefficient travel and non-zero-coefficient travel and makes variable-length coding on them, and as coding the value information, combines the continuous plus or minus 1 and the values or absolute values of the back coefficients with the absolute values greater than 1 into two-dimensional pairs of the travel of plus or minus 1 and the said value or absolute value, and makes the variable-length coding on them, or uses 1 bit to indicate if the absolute values of the string of non-zero coefficients are all 1. The invention makes further use of the statistical properties in the position information and the value information of the non-zero coefficients, thus obtaining a prior compressing performance to that of the existing entropy coding method.

The invention is related to entropy coding / decoding of transform coefficient data in video compression systems. For entropy coding coefficients representing a block in a video image, a preferred embodiment of the present invention introduces events combining the position of the last non-zero coefficient in the block with whether the absolute value is greater than 1. Further, no information from outside the macroblock is used to decide what VLC to use. Coefficients are typically coded by starting in a Run-mode and continuing in Level-mode when the first coefficient with absolute value >1 is found.

A method of image coding wherein an image is divided into blocks having a plurality of pixels. A transform coding operation is performed on a block of pixels to produce a corresponding block of transform coefficient values, which is scanned to produce a scanned array of coefficient values represented by a plurality of number pairs having a first and a second number. The first and second numbers are assigned to one of a plurality of contexts (14) representative of the number pairs. The first number of a number pair is assigned to a context based on a first number of another number pair. Alternatively, the second number of a number pair is assigned to a context based on the first number of the number pair (16). Furthermore, a number indicative of the number of non-zero coefficient values in the block of transform coefficient values is determined and assigned to a context.

The invention discloses an image coding method and system based on dictionary learning. The method comprises the steps that image blocks are randomly selected from an image sample to carry out sparse representation based on batch processing, and a dictionary matrix is solved through an alternative optimization method; according to the obtained dictionary matrix, the sparse representation processed in batch is utilized to carry out sparse representation on the image blocks to be coded, and quantification entropy coding is carried out on nonzero coefficients in sparse coefficient vectors; decoding and sparse reconstruction are carried out on the coded image at a decoding end according to the obtained dictionary matrix. The system comprises a dictionary learning module of sparse representation based on batch processing, an image coding module and an image decoding module. According to the method and system, the problem that the training sample in dictionary learning is large in scale is resolved, meanwhile reconstruction errors of the image sample are reduced, and rate-distortion performance of image compression is significantly promoted.

The invention discloses a parallel non-zero coefficient context modeling method for binary arithmetic coding, and relates to a context modeling technology for video coding. The method is proposed to solve the problem that the data throughput rate of a coding system is reduced because the conventional binary arithmetic coding generates a data dependency relationship on the context in the context modeling process of non-zero coefficients. The method comprises the following steps of: 1, defining the number of coefficients and non-zero coefficients in a transform quantification block; 2, performing binarization on the non-zero coefficients to obtain a bin sequence; 3, performing context modeling on a first context according to the position information of the non-zero coefficients and the number of the non-zero coefficients in the transform quantification block; 4, calculating the probability distribution of the non-zero coefficients, the absolute value of which is abs (Li) in the first context value; 5, subtracting 1 from the absolute value of Li, and performing binarization; and 6, performing context modeling by using the equal probability distribution. By using the method, the context modeling processes of different non-zero coefficients can be simultaneously performed, and parallel execution of multiple context modeling processes in the coding process is realized.

The invention discloses a video image coding / decoding method based on geometric partitioning. The coding method comprises the following steps: firstly, a video image is partitioned into a plurality of rectangular coding blocks, and the rectangular coding blocks are geometrically partitioned; secondly, motion estimation is carried out for each geometrically partitioned irregular block, each irregular block obtains an own residual block, and the residual blocks corresponding to the irregular blocks are combined to a rectangular residual block; thirdly, the coordinates of pixels in the rectangular residual block are rearranged by utilizing geometrically partitioned boundary direction information; and lastly, two-dimensional orthogonal transformation is carried out for the rearranged rectangular residual block, a transformation coefficient is quantized and entropy coding is processed, in addition, geometrically partitioned information and rearranged information are incorporated into a code flow. The decoding process and the coding process are inverse. Since the geometrically partitioned boundary direction information is utilized and the pixels of the rectangular residual block are rearranged when the coding is processed, so that after orthogonal transformation, a high-frequency nonzero coefficient is reduced, and compression efficiency is improved.

The disclosed selecting method for quantizing matrix comprises: A. determining the position of current coding block, quantizing the block with high-precise quantizing matrix if current block as the initial one, or else going to next step; B. obtaining the non-zero coefficient number predicted value (NumCoeffPred) for current block; and C. comparing the predicted value with threshold, and selecting the high- / low-precise quantizing matrix if over / lower the threshold. This invention saves coding bit, and improves coding efficiency.

Further compression of data allowing economical storage of data for extended periods of time in high-speed access memory is performed in reduced time by performing further compression during transcoding in the transform domain and without restoring image data to its original image data form. The reduction in processing time is achieved by exploiting the large number of zero-valued quantization transform coefficients and not changing quantized transform coefficients at zig-zag scan positions where non-zero coefficients are rare during range reduction of the entropy decoded quantized transformed data. The range can be restored by computation or estimation of an altered quantization table which is stored with the further compressed a quantization values. Further advantages accrue from use of JPEG packed format for the data during transcoding.

This disclosure describes techniques for quantizing coefficients of a video block for a video coding process that supports context-based adaptive binary arithmetic coding (CABAC). A method may comprise generating a plurality of sets of quantized coefficients for the video block, wherein each of the sets of quantized coefficients is defined based on an assumption that a particular one of the coefficients of the video block corresponds to a last non-zero coefficient of the video block, estimating a plurality of costs associated with coding the video block via the CABAC, wherein the estimated costs are based respectively on different ones of the sets of quantized coefficients, and selecting a given one of the sets of quantized coefficients associated with a lowest cost.

A block transform-based digital media codec more efficiently encodes transform coefficients by jointly-coding non-zero coefficients along with succeeding runs of zero-value coefficients. When a non-zero coefficient is the last in its block, a last indicator is substituted for the run value in the symbol for that coefficient. Initial non-zero coefficients are indicated in a special symbol which jointly-codes the non-zero coefficient along with initial and subsequent runs of zeroes. The codec allows for multiple coding contexts by recognizing breaks in runs of non-zero coefficients and coding non-zero coefficients on either side of such a break separately. The codec also reduces code table size by indicating in each symbol whether a non-zero coefficient has absolute value greater than 1 and whether runs of zeros have positive value, and separately encodes the level of the coefficients and the length of the runs outside of the symbols.

A convolution engine, such as a convolution neural network, operates efficiently with respect to sparse kernels by implementing zero skipping. An input tile is loaded and accumulated sums are calculated for the input tile for non-zero coefficients by shifting the tile according to a row and column index of the coefficient in the kernel. Each coefficient is applied individually to tile and the result written to an accumulation buffer before moving to the next non-zero coefficient. A 3D or 4D convolution may be implemented in this manner with separate regions of the accumulation buffer storing accumulated sums for different indexes along one dimension. Images are completely processed and results for each image are stored in the accumulation buffer before moving to the next image.

An image data decoding apparatus according to an embodiment of the invention includes: a variable length coder / decoder for decoding compressed image data to output a quantization coefficient; a non-zero coefficient detector for examining whether or not the quantization coefficient includes a non-zero coefficient for each examination unit; a quantization coefficient FIFO for storing, if the examination unit includes one or more non-zero coefficients, the quantization coefficient of the examination unit as a significant coefficient; a significant coefficient flag FIFO for storing a significant coefficient flag that indicates whether or not the examination unit is a significant coefficient, and a zero coefficient generator for reconstructing the quantization coefficient based on the significant coefficient flag and significant coefficient.

This invention discloses a method for an adaptive variable-length decoding based on the context controlled by a controller including decoding five grammar elements: beginning from coeff token, numbers of total coeff of non-zero factors and trailing ones of trailing one, decoding trailing one sign, the symbol of trailing one, level, the amplitude of the left non-zero coefficient, decoding of total zeros, the number of zero before the last non-zero coefficient, run before decoding number of zero from the last non-zero coefficient to the front non-zero coefficient, in which, coeff token decoding applies a multi-ID number distribution method based on the packet rule or a storage conversion method and decoding of total zeros applies the multi-ID number distribution method based on the packet rule or a code word merging method based on the relation of the code words after packet.

The invention discloses a method for implementing coding, comprising the following steps: determining the number of nonzero coefficients in current block, and coding the number; scanning the coefficients of current block to form binary number pairs each of which is composed of the number of zero coefficients before a nonzero coefficient and the nonzero coefficient, mapping the nonzero coefficient to a level and a residual amplitude according to mapping relation; and coding the number of zero coefficients before the nonzero coefficient, the serial number corresponding to the level and the residual amplitude respectively. According to the invention, the distribution law between a Run and the level as well as the correlation between the Run and the number of nonzero coefficients are effectively used, which lowers coding bit rate, saving bit stream and improves coding efficiency. The invention also discloses a device for implementing coding, a method and a device for implementing decoding.

The present invention provides an image encoding device which does not necessitate reference of a quantization parameter between consecutive macroblocks across a parallel processing area boundary without forming slices. The image encoding device encodes a macroblock of an encoding target image by parallel processing sequentially from the top of a parallel processing area, and possesses an encoding element for every parallel processing area. When all the quantized orthogonally-transformed coefficients of a top macroblock of the parallel processing area are zero, the encoding element adds a non-zero coefficient to a part of the coefficients, making the coefficients non-zero. Accordingly, generation of a skip macroblock in the top macroblock of each parallel processing area is suppressed. Since slice formation is not necessary, the prediction over a parallel processing area boundary is applied, and encoding efficiency improves. Errors are not generated in decoding and the decoded image quality does not deteriorate.

This invention discloses one method to strengthen picture element area codes, which comprises the following steps: a, decoding the input codes of MPEG-2 input code flow; outputting decoding information besides the video data of YUV form; if each frame of image coding mode, each macro coding mode; then making full use of MPEG-2 decoding to get output information and re-coding the YUV form visual data for AVS data; finally coding, and considering the characteristics of MPEG-2 and AV codes with quality and speed.

The invention discloses a decoding method of the H.264 context self-adapting coding residual error coefficient. The method comprises following steps: firstly decoding the quantity of the nonzero coefficient and a binary group of the trailing coefficient quantity; then decoding the sign digit of the trailing coefficient; then decoding the extent value of the nonzero coefficient of the trailing coefficient; finally decoding the quantity of the previous zero number of the last nonzero coefficient and the quantity of the previous zero number of each nonzero coefficient. The binary group decoding of the nonzero coefficient and a binary group of the trailing coefficient can confirm the length of the code information according to the three-dimensional code list and the predetermined bit digits, further confirming the content of the binary group; the quantity of the nonzero coefficient and a binary group of the trailing coefficient quantity and the extent value of the nonzero coefficient use the quantity obtained by the method of two bytes of the prior zero. The method can enhance the decoding speed of the CAVLC code flow.

The present invention enables a circuit configuration to be simplified. An image encoding device (100) receives an input image (91) that is line-scanned in units of MBs as scan blocks, each constituted by a plurality of 4×4 blocks, while being scanned in units of 4×4 blocks as encoding blocks, each constituted by a plurality of pixels. The image encoding device (100) selects a VLC (variable length) table corresponding to a table selection index nC, which is an average value of the numbers nB and nA of nonzero coefficients as encoding values on which variable-length encoding is performed in upper-adjoining and left-adjoining 4×4 blocks of a 4×4 block as an encoding target in quantization coefficients as encoding target data based on the input image (91). At this time, in a case where the 4×4 block as an encoding target is at an upper end of an MB, the image encoding device (100) equates the number nB of nonzero coefficients of the upper-adjoining 4×4 block with the number nA of nonzero coefficients of the left-adjoining 4×4 block. The image encoding device (100) performs variable-length encoding on the number of nonzero coefficients of a 4×4 block as an encoding target in quantization coefficients using the selected VLC table.