BR112017023403B1 - METHOD AND APPARATUS FOR ENTROPY CODING OF IMAGE OR VIDEO DATA - Google Patents

Abstract

METHOD AND APPARATUS FOR ENTROPY CODING OF IMAGE OR VIDEO DATA. An entropy coding method and apparatus for a video encoder or decoder using Context-Based Binary Arithmetic Encoder (CABAC) based on multiple tables are described. In one embodiment, a current bin of binary data of a current encoding symbol is encoded or decoded according to a probability of a binary value of the current bin and the probability of the binary value is updated according to the binary value of the current bin to a following bin using multi-parameter probability models. Each multi-parameter probability model is updated using at least one look-up table with the individual probability state set as a table index to access the contents of this at least one look-up table. In another embodiment, the range update is calculated for a range based on the average value of the range interval.

Description

CROSS REFERENCE TO RELATED ORDERS

[001] The present invention claims priority to U.S. provisional patent application Serial No. 62/163,473, filed on May 19, 2015, U.S. patent application Serial No. 62/214,129, filed on September 3, 2015, and U.S. provisional patent application Serial No. 62/322,306, filed April 14, 2016. The U.S. provisional patent applications are hereby incorporated in their entirety herein by reference.

FIELD OF INVENTION

[002] The present invention relates to entropy coding techniques for image coding and video coding. More especially, the present invention relates to a Context-Based Binary Arithmetic Coder (CABAC) based on multiple tables for image coding and video coding.

FUNDAMENTALS AND RELATED PREVIOUS TECHNIQUE

[003] Arithmetic coding is known as one of the efficient data compression methods and is widely used in coding standards, including JBIG, JPEG2000, H.264/AVC and High Efficiency Video Coding (HEVC). In H.264/AVC and HEVC Test Model Version 16.0 (HM-16.0), context-based adaptive binary arithmetic coding is adopted as the entropy coding tool in the video coding system.

[004] As shown in Figure 1, CABAC consists of three parts: binarization unit 110, context modeling unit 120, and binary arithmetic coding unit 130. In the binarization step, each syntactic element is uniquely mapped to a binary string (bin or bins). Within the context modeling step, a probability model is selected for each bin. The corresponding probability model may depend on previously encoded syntactic elements, bin indices, and collateral information. After binarization and context template assignment, a bin value along with its associated template is passed to the binary arithmetic coding engine.

[005] Binary arithmetic coding is a recurring interval subdivision procedure. The output bitstream is the indicator of the final probability of the coding interval. The coding range probability, T, is specified by the range and the lower bound of the coding range (designated as “low” in the discussion that follows). The range is the possible scope of the encoding range. Depending on whether the current symbol is the most likely symbol (MPS) or the least likely symbol (LPS), the next encoding range is updated as one of the two subranges accordingly as shown in eq. (1) and in eq (2). where RangeLPS is the estimated range when LPS is encoded.

[006] Figure 2 illustrates the idea of binary arithmetic coding. Initially, the probability range (i.e., range0) is 1 and the low limit (i.e., low0) is 0 as indicated by probability scale 210. If the first symbol is an MPS symbol, an indicator at the bottom of the scale probability 210 can be used to signal the event of an MPS symbol. Range1 is used as the probability scale 220 for processing the next symbol. Again, the 220 probability scale is divided into two parts for MPS and LPS, respectively. If the second symbol is an LPS symbol, RangeLPS1 will be selected as the 230 probability scale for the next symbol. Whenever a new symbol is encoded, the corresponding range becomes smaller. When a range becomes too small, the range can be renormalized to form a 240 probability scale with a larger range.

[007] In modern arithmetic coding, probability updating is often performed according to a template. One method is described, for example, by Marpe et al., in a technical publication (“Context-Based Adaptive Binary Arithmetic Coding in the H 264/AVC Video Compression Standard”, IEEE Transactions on Circuits and Systems for Video Technology, Vol. 13, No. 7, pp 620-636, July 2003), where the following formula is used: In the above equation, y equals 0 if the current symbol is the most likely symbol (MPS), otherwise y equals 1. This formula provides an estimated value for the probability of the least likely symbol (LPS). The weighting a is derived according to the following equation where min_prob corresponds to the minimum probability of the least likely CABAC symbol and state_number corresponds to the number of context states for probability value estimation.

[008] For HEVC CABAC, there are 64 probability states. The min_prob is 0.01875, and the state_number is 63. Each state has a probability value indicating the probability for selecting the LPS. The 64 representative probability values, pc and [0.01875,0.5] , were derived for the LPS by the following recursive equation:

[009] The RangeLPS value of a state a is derived from the following equation:

[0010] To reduce the computations required to derive RangeLPS, the RangeLPS result of each range value can be pre-calculated and stored in a look-up table (LUT). In H.264/AVC and HEVC, a 4-column table of pre-calculated RangeLPS is adopted to reduce the table size as shown in Table 1. The range is divided into four segments. In each segment, the RangeLPS of each state of probability a (pa) is pre-defined. In other words, the RangeLPS of a state of probability a is quantified into four values. The selected RangeLPS depends on the segment the range belongs to.

[0011] In JCTVC-F254 (Alshin et al., “Multi-parameter probability up-date for CABAC”, Joint Collaborative Team on Video Coding (JCT-VC) of ITU-T SG16 WP3 and ISO/IEC JTC1/SC29/ WG11, 6th Meeting: Turin, IT, July 14-22, 2011, Document JCTVC-F254), Alshin, et al., release a multi-parameter probability update for the HEVC standard CABAC. The parameter N = 1/(1-a) is an approximate measure for the number of previously encoded bins (i.e., “window size”) that have a significant influence on the current bin. The choice of parameter N determines the sensitivity of the model. A sensitive system will react quickly to a real change. On the other hand, a less sensitive model will not react to noise and random errors. Both properties are useful but contradictory. Consequently, Alshin et al., describe a method for calculating several values with different ai simultaneously: and uses weighted average as the probability prediction of the next bin: where ßi is a weight factor associated with ai.

[0012] Instead of the state transition lookup tables (m_aucNextStateMPS and m_aucNextStateLPS) used in CABAC of the AVC standard for updating the state and its corresponding probability, Alshin, et al., uses explicit calculation with a multiplication-free formula to update the probability. Assuming that the probability pi is represented by the integer Pi from 0 to 2k (i.e., pi= Pi/ 2k) and that ai is represented by 1 over a power of the number two (i.e., ai=1/2Mi), the Multiplication-free formula for updating the probability can be derived as follows:

[0013] This formula predicts the probability that the next bin will be “1” where Y = 2k if the last encoding bin is “1”, Y = 0 if the last encoding bin is “0”, and “> >Mi” corresponds to the shift to the right of an Mi bit operation.

[0014] To maintain the balance between increased complexity and increased performance, the linear combination for probability estimation is considered to consist of two parameters only:

[0015] The floating point value that corresponds to the probability for CABAC stroke is always less than or equal to 1/2. If the probability exceeds this limit, LPS becomes MPS to keep the probability within the range mentioned above. It is necessary to switch MPS/LPS when the MPS/LPS probability is greater than 0.5. Alshin,et al.,proposed to increase the permissible level of probability,(in terms of floating point values) up to 1 to avoid MPS/LPS,switching. Therefore, a look-up table (LUT) is derived for storing RangeOne or RangeZero.

[0016] In VCEG-AZ07 (Chen, et al., “Further improvements to HMKTA-1.0”, ITU-T Video Coding Experts Group (VCEG) meeting, Warsaw, Poland, IT, June 19-26, 2015, Document: VCEG-AZ07), Chen, et al., proposed to use a single-parameter CABAC. The probability derivation is the same as that of JCTVC-F254, which uses eq. (9) to derive the probability of it being one or zero. For each context, only one update speed is used, which means for each context only one N is used. However, different contexts may use different Ns. The range for N varies from 4 to 7, and a 2-bit variable is used to indicate the probability update speed for a specific context model. The value of N is determined on the encoder side and signaled in the bitstream.

[0017] In JCTVC-F254 and VCEG-AZ07, the RangeOne or RangeZero LUT is a table of 64 columns by 512 rows. The LUT launch consists of the current range and the current probability. The valid range of the current range is 256 to 510. The current range is divided into 64 sections, where each section contains 4 current range values (such as, for example, 256 to 259, 260 to 263, etc.). The track section index can be derived by:

[0018] The valid range of the current probability (P) extends from 0 to 2k - 1. In JCTVC-F254 and VCEG-AZ07, the value of k is 15. The current probability is divided into 512 sections, each section containing 64 values current probability (such as, for example, 0 to 63, 64 to 27, etc.). The probability section index can be derived by:

[0019] The RangeOne value can be derived by consulting the table, for example:

[0020] Each value in table RangeOne (tableRangeOne) is derived by: where MinRange is the lowest range value of the derived rangeIdx. clip3(X, Y, Z) is for trimming the Z value within the range of X to Y. Round is for rounding the value up to an integer.

[0021] The range section for rangeIdx = 0, for example, is 256 to 259, the MinRange value is 256. The MinRange value can be derived by:

[0022] The value of M is the maximum value of (probIdx + 1). In JCTVC-F254 and VCEG-AZ07, for example, the M value is 512. Table 2 shows the lookup table published in JCTVC-F254, which consists of 64 columns for range values and 512 entries. For each release the range value is represented by 9 bits. Table 2

[0023] Two in-circuit filters are included in the H.265/HEVC video coding standard. They consist of deblocking filter and sample adaptive offset (SAO). The deblocking filter can reduce blocking artifacts that are caused by quantification error. SAO can also improve video quality by applying offset values to the classified samples. Prior to the HEVC Test Model 7 (HM-7) test model, another in-loop filtering technique called adaptive in-loop filter (ALF) was also included. ALF uses Wiener filtration techniques to derive filtration coefficients. A multitude of filters are encoded according to different pictorial regions. Filtering coefficients are encoded in adaptation parameter sets (APS), and on/off control flags are encoded using CTU-level (coding tree unit level) syntactic elements.

[0024] It is obvious that filtering coefficients are the main surplus in bitrate when encoding the syntactic elements of ALF. Generally, the texture characteristics of the neighboring coding block are very similar to those of the current coding block. Therefore, the neighboring coding block filter can be directly used for the current coding block to save bitrate surplus. Since two neighboring blocks apply the same filter coefficients in this case, this coding method is also called filter fusion. A priority-based block filter fusion scheme was also disclosed. The first step consists of choosing the maximum number of N candidates from N predefined neighboring blocks. The second step consists of selecting a filter among N candidates and encoding its filter index into the bitstream. Next, a method is described to further improve the performance of the priority-based block filter fusion scheme.

[0025] When a multiplicity of filters is sustained in ALF, in addition to dividing an image into different regions, some pixel-based or block-based classification methods are also present before HM-7. An ALF technique has been described, for example, to calculate the Summation-Modified Laplacian Measure (SLM) of each pixel. Pixels having the same SLM value will be filtered by a filter. As shown in Figure 3, each square indicates a pixel, and SLMn pixels are filtered by a filter, where n can be 1, 2 or 3 in this example. In this case the SML is treated as a type of pixel classification rule (PCR). For a block-based classification method, the first step is analogous to pixel-based classification to calculate the characteristic of each pixel in a block. The second step is to calculate the property of a block based on the characteristics of all pixels in a block.

BRIEF SUMMARY OF THE INVENTION

[0026] A method and apparatus for entropy coding image and video data to an image or video encoder or decoder using Context-Based Adaptive Binary Arithmetic Encoder (CABAC) are described. In one embodiment, a current bin of a binary data set of a current encoding symbol is encoded or decoded depending on a probability of the binary value of the current bin and the probability of the binary value is updated according to the binary value of the current bin. for a following bin using multi-parameter probability models. The probability of the binary value of the current bin is generated from one or more previously encoded symbols before the current encoding symbol. Each multi-parameter probability model is updated using an individual set of probability states associated with a corresponding parameter. Specifically, each multi-parameter probability model is updated using at least one look-up table with the individual probability state set as a table index to access the contents of the at least one look-up table.

[0027] In one example, the lookup table comprises an LPS range table (least likely symbol), including the LPS range table includes the predetermined LPS range for a given probability state and a current range. The LPS range table may include the predetermined LPS range for the given probability state and a quantified current range to reduce the size of the table. The LPS range table can store range values for a reduced number of the individual set of probability states, and the reduced number of the individual set of probability states is selected by uniform conservation of one probability state from each M probability states. and M is a positive integer greater than 1. The value M corresponds, for example, to 2, 4, or 8. The LPS range table can store range values for a reduced number of the individual set of probability states and the reduced number of the individual set of probability states is selected by non-uniform conservation of the individual set of probability states.

[0028] The lookup table may comprise an LPS next state table (least likely symbol) or an MPS next state table (most likely symbol) wherein the LPS next state table or the next MPS state table MPS includes a following LPS probability state for each current LPS probability state or a following MPS probability state for each current MPS probability state. The LPS next states table may store the LPS next states for a reduced number of the LPS probability states or the MPS next states table stores the MPS next states for a reduced number of the MPS probability states.

[0029] Multi-parameter probability models can correspond to two-parameter probability models using a first parameter and a second parameter, and the first parameter is derived based on the second parameter. The first parameter can be equal to the second parameter raised to a power of M and M is an integer greater than 1. A probability of the binary value can be derived from new individual probabilities updated according to multi-parameter probability models. A derivation of the final probability based on the probabilities of two respective probability states associated with two probability parameters is also described.

[0030] Another method of entropy coding image and video data in an image or video encoder or decoder is also described. The current bin of a set of binary data of a current encoding symbol is encoded or decoded according to a probability of a binary value of the current bin, being the probability of the binary value of the current bin generated from one or more encoded symbols before the current encoding symbol. The probability of the binary value is then updated according to the binary value of the current bin for a next bin; and encoding or decoding the current bin using range One or range Zero values derived from at least one range lookup table. A Range Lesser Half (RangeSH), Range One values, or Range Zero values are derived for a given range interval based on the given range range and a given probability of the binary value.

[0031] The at least one range lookup table may comprise a Zero range table or a One range table. The One range value may be derived for an average range value of the given range range, and a range value average probability of given range interval from probability range Zero or probability range One. The range value of One can be derived for an average range value of the given range range and a maximum probability value of the given range interval from the range of probability Zero or probability range One. The at least one range lookup table may only include range values for the probability range Zero or probability range One between 0.0 and 0.5. The range value of the Zero probability range or the One probability range between 0.5 and 1.0 can be derived by (current range - the range values for the One probability range or the Zero probability range between 0.0 and 0.5). The at least one range lookup table may include range values for the Zero probability range or One probability range between 0.0 and 1.0, the range values for the Zero probability range or the One probability range being One between 0.5 and 1.0 are reflected from the range values for the Zero probability range or the One probability range between 0.0 and 0.5.

BRIEF DESCRIPTION OF THE DRAWINGS

[0032] Figure 1 illustrates a basic structure of context-based adaptive binary arithmetic coding (CABAC).

[0033] Figure 2 illustrates a conception of binary arithmetic coding, in which initially the range of probabilities is 1 (i.e. range0) and the low limit (i.e. low0) is 0 as indicated by a probability scale.

[0034] Figure 3 illustrates an example of Summation-Modified Laplacian Measure (SLM) of each pixel, each square indicating a pixel and SLMn pixels are filtered by a filter and n can be 1, 2 or 3 in this example.

[0035] Figure 4 illustrates an example of table-based two-parameter context-based adaptive binary arithmetic coding (CABAC), where one parameter provides a faster update speed and the other parameter provides a slower update speed.

[0036] Figure 5 illustrates an exemplary flowchart for a context-based adaptive binary arithmetic coding (CABAC) based on multiple tables in accordance with an embodiment of the present invention.

[0037] Figure 6 illustrates an exemplary flowchart for another context-based adaptive binary arithmetic coding (CABAC) based on multiple tables in accordance with an embodiment of the present invention.

DETAILED DESCRIPTION OF THE INVENTION

[0038] The description that follows is the best contemplated way of implementing the invention. The description is given for the purpose of illustrating the general principles of the invention and should not be considered to have a limiting meaning. The scope of the invention is best determined by reference to the appended claims.

[0039] In JCTVC-F254 and VCEG-AZ07, instead of storing the probability state index, the actual probability of each context is stored in a 16-bit or 32-bit buffer. Compared to the 7-bit state index in HEVC, implementation costs increase substantially. To derive the range from RangeOne or RangeZero, a 9-bit*64 columns*512 entries (i.e. 294912 bits) look-up table is used. The size of the lookup table is quite large for a parser. Accordingly, in the present invention, a multiple table-based CABAC encoding is disclosed. Using the formula in eq. (4) or in eq. (19), one or a multiplicity of values of a are derived. Once a is derived, other values of a can be derived using eq. (20):

[0040] The value of N, for example, can be 7 and M can be 16. Consequently, a2 can be 1 - (1/128) and a1 can be (a2)16.

[0041] Using the derivative and eq. 3 or eq. 21 (eq. 3 multiplied), the probability of LPS is also derived:

[0042] The multiplicity of values of a for deriving the probability states mentioned above are called multiple parameters or multiple model parameters in this description. As shown above, these probability states correspond to a set of predefined values associated with the LPS or MPS probability. However, other parameters, such as N and M in the example above, can also be used directly or indirectly as probability model parameters. Although the shape of the probability state is described using eq. (21) and the parameter forms are described in eqs. (19) and (20), it should be noted that other forms of equivalent parameters are also used in this field (1-a), (1-a1) and (1-a2) have already been used to replace a, a1 and a2 in the eq. (19) to (21). It should be understood that these different forms are equivalent and can be used interchangeably.

[0043] Using the new a and the new p, the new LPS range tables for a1 and a2 can be derived as shown in Table 3 (for a1) and Tables 4a to 4h (for a2) respectively. Table 3

[0044] Figure 4 illustrates an example of table-based two-parameter Context-Based Adaptive Binary Arithmetic Coding (CABAC), where one parameter (i.e., N1 corresponding to a1) provides a faster update speed and the other parameter (i.e. N2, corresponding to a2) provides a slower update speed.

[0045] For each context, it has a multiplicity of probability states associated with a multiplicity of parameters. It can have two associated probability states, for example, with two parameters for each context: one for a1 and another for a2. The states (state a and state b, for example) in each context are updated independently. For encoding/decoding a bin, one state can be updated to the LPS state and the other can be updated to the MPS state. When deriving the LPS range for coding (RLPSC), the LPS range for each state can be derived by consulting the table. If the MPS of two states is the same, the RLPSC is the average of the LPS range of state a (RLPS_a) and the LPS range of state b (RLPS_b). Otherwise, the average range of bin-0 and the average range of bin-1 are derived. The RLPSC is the average range with the lowest value. If RLPS_a is smaller, for example, RLPSC is equal to average RLPS_a and range_MPS_b. If RLPS_b is smaller, RLPSC will be equal to average RLPS_b and range_MPS_a. Range_MPS_x is equal to (range - RLPS_x), where x corresponds to a or b. The averaging operation between value A and value B can be implemented using the right shift operation, such as ((A+B)>> 1) or ((A+B+1)>>1 ). The RLPSC, RMPSC (MPS range for coding), and MPS can be derived according to the process in the table below: Table 5

[0046] The derived LPS range for coding and the MPS range for derived coding, which is equivalent to (range - RLPSC) can be used for HEVC CABAC. Table-based multi-parameter CABAC as shown above can substantially reduce the size of the look-up table (LUT).

[0047] In CABAC based on multiple tables, the MPS values or the LPS values associated with the probability states for a context may be different. As mentioned above, the two probability states for the two-table case can have different MPS or LPS values. Therefore, instead of dealing with the probability model associated with MPS and LPS, it is also possible to deal with bin values 0 and 1. Therefore, another method is to use the LPS table to derive RangeOne or RangeZero for each probability state, with RangeOne being the range value for bin value 1 and RangeZero being the range value for bin value 0. The mean of RangeOne or RangeZero can be derived by calculating the mean of RangeOne and of RangeZero, respectively. The RangeOne for coding (ROFC) and the RangeZero for coding (RZFC) can be derived by a process as shown in Tables 6 through 9. Table 6

[0048] ROFC and RZFC can be used for CABAC. Table-based multi-parameter CABAC as described above can substantially reduce LUT size.

[0049] To further reduce the LUT size, the LUT can be reduced by its sampling rate. The LUT for a1, for example, may have its sample rate reduced. The LUT with the reduced sample rate can be an LPS transition sample and/or an LPS track table. Two types of sample rate reduction methods are shown below.

[0050] Uniform quantification: - Compression of M:1 by storing the states 0, M, 2M, 3M, ..., where M = 2, 4, 8. - For the LPS transition, next_LPS_state(K) = next_LPS_state (K/M) + K%M or next_LPS_state(K) = next_LPS_state (K/M) + K%M except for K = 0, or K = 0, 1.

[0051] Non-uniform quantification: - No compression for states in [0,N/4-1], - 2:1 compression for states in [N/4, N/2- 1], - compression of 4: 1 for the states in [N/2, N-1], and - N can be 512

[0052] Table 10 illustrates, for example, an example of the LPS range LUT for a1, with the compression ratio M = 8. For a state K, its RangeLPS is LUT(K/M).

[0053] As a1 and a2 are related to each other by a1 = (a2)16, the table from a2 can be reused for a1. State S1 in a1 is the same as state S1*16 in a2. State 1 in a1, for example, is the same as state 16 in a2, and state 2 in a1 is the same as state 32 in a2.

[0054] Table 11 illustrates an example of the following LPS state for a1, where “-1” in the table means the change to MPS and the state is set to 0 and “-2” means the change to MPS and Table 11

[0055] Table 12 illustrates an example of the following LPS state for a2 with the compression ratio M equal to 8. For a state K, its next LPS state is equal to (next_LPS_state(K/M) + K%M) . For state 0, your next LPS state can be -1 or -2, where “-1” means change to MPS and the state is set to 0, and “-2” means change to MPS and the state is adjusted to 1. Table 12

[0056] For state initialization, the initial state derivation used by HEVC can be reused. However, a look-up table for mapping the initial state (probability) in HEVC to the nearest initial state (probability) is used in accordance with the table-based multi-parameter CABAC of the present invention. Table 13 illustrates an example of the initial state mapping table for a1 and a2. Table 13

[0057] In the present invention, we also propose the use of more than one parameter (more than one a, for example) for CABAC coding. For each a, it has its states RangeLPS table (or RangeOne table, or RangeZero table), next MPS table, and next LPS table. For each context he can choose to use a single a or two a. If two a are used, the methods mentioned above can be used. Some syntaxes may use a single a, and some syntaxes may use two a's. This secondary information (such as information to indicate syntax using one a or two a and to identify which a is, for example) may be predefined or signaled in the bitstream. The coefficient related syntaxes, for example, may use a single a, and other syntaxes may use two a. In another example, coefficient-related syntaxes may use two a, and other syntaxes may use a single a. These values of a can be derived using eqs. 4, 19 and 20, for example, a5 can be 1-(1/128); a1 can be (a5)16; a2 can be (a5)8; a3 can be (a5)4; a4 can be (a5)2. The RangeLPS tables for others are also predefined. Each context stores information regarding what it is used for and the current state.

[0058] In CABAC, the valid range value ranges from 256 to 510. For RangeLPS derivation, the number of columns required depends on the range resolution. In Table 3 and Tables 4a-4h, for example, the range is divided into four parts: 256 to 319, 320 to 383, 384 to 447, and 448 to 510. The mid_value of each range part can be derived appropriately, such as 288, 352, 416, and 480 respectively. For the first state (first release) in each column, the probability is (0.5 * mid_value). The second state in each column is the first state multiplied by a. The following states in each column consist of the previous state multiplied by a. For RangeLPS derivation, the mid_value can be changed to a value equal to or greater than the minimum range value (256, 320, 384, and 448, for example) and equal to or less than the maximum range value (319, 383, 447 , and 510, for example).

[0059] In JCTVC-F254 and VCEG-AZ07, the RangeOne table covers probability from 0.0 to 1.0. However, this makes the LUT too large to implement in terms of hardware cost. The LUT is 144 times larger than the LUT of HEVC. Furthermore, since each RangeOne or RangeZero launch value is derived from the MinRange (eq. 18), encoding efficiency will drop dramatically if the LUT with the reduced sample rate is used.

[0060] Therefore a method is disclosed in the present invention for storing the probability range from 0.0 to 0.5 only, which is called the lower half of the range (RangeSH). Values in the other half of the range can be derived using (Range - RangeSH). The number of lines defines the resolution of the probabilities. A RangeSH table, for example, with 64 rows can be designed for a probability range of 0.5 to 0.0. Each row represents the RangeLPS for the 1/64 probability range. The RangeSH value is derived by ((Range A) * (Prob B)). Table 14 illustrates an exemplary RangeSH table with 4 columns and 64 rows. The first row represents the RangeSH for probability from 63/128 to 64/128 in four different range sections. In Table 14, Range A corresponds to Range Mid and Prob B corresponds to Prob Max. The RangeSH value is derived by ((Range Mid) * (Prob Max)). Table 14

[0061] Table 15 illustrates another exemplary derivation method for RangeSH, which is derived from ((Range Mid) * ((Prob Max+ Prob Min)/2)). Table 15

[0062] Table 16 illustrates another exemplary value derivation method, in which RangeSH is derived by ((Range Mid) * ((Prob Max + Prob Min)/2)) and the minimum value is trimmed to 3. In JCTVC-F254 and VCEG-AZ07, if the probability of one is greater than 0.5 (0.64, for example), it means that the probability of zero is 0.36. The value 0.36 (on the 18th row) will be used to find the range for RangeZero. RangeOne can be derived by (range - RangeZero). Table 16

[0063] In an embodiment for deriving RangeOne (or RangeZero), the probLPS value can be derived using the expression: probLPS = (P >= 2k-1)? 2k-1-P:P for a k-bit probability (2k>P> 0). The expression “x ? y : z” represents a logical operation, and if x is TRUE or not equal to 0, it evaluates to the value of y; otherwise, it evaluates to the value of z. The probIdx value can be derived as (probLPS >> (k-n-1)), where the RangeSH table has 2n rows. The value of rangeIdx is derived as (range>>(8-m)) - (256>>m), ((range-256)>>(8-m)), or ((range>>(8-m ))&(2 m -1)), where the RangeSH table has 2m columns. RangeSH is determined from the RangeSH [probIdx] [rangeIdx] table. If P is equal to or greater than 2k-1 (or the kth bit of P is 1), RangeOne is equal to (range - RangeSH) and RangeZero is equal to RangeSH. Otherwise (i.e. P less than 2k-1), RangeOne is equal to RangeSH and RangeZero is equal to (range - RangeSH).

[0064] In the example of JCTVC-F254 and VCEG-AZ07, k is 15, the probLPS value is determined from the expression: probLPS = ((P>= 16384) ? 32767-P: P), probIdx is equal to ( probLPS>>8), rangeIdx is equal to (Range>>6) & 3. If P is equal to or greater than 16384, RangeOne is equal to (range - RangeSH) and RangeZero is equal to RangeSH. Otherwise (that is, if P is less than 16384), RangeOne is equal to RangeSH and RangeZero is equal to (range - RangeSH).

[0065] Note that, as (2k - 1) all consist of a binary representation, then (2k - 1 - P) exists only to perform the inverse of binary logic (bitwise) for the k bits of LSB (the bit less significant). In hardware implementation, exclusive binary logic or (XOR) for the k-order bit of P and the 0-order (k) bits of P to derive the probLPS value.

[0066] In another embodiment, the RangeSH table is duplicated to reduce computing complexity. Table 17 illustrates an example of the table reflected from Table 16. Throws are reflected at the boundary between the probIdx value 63 and 64. Using this type of RangeSH table, the probIdx value can be derived by probIdx = (P >> (kn)) directly, where the RangeSH table has 2n rows. In the example of JCTVC-F254 and VCEG-AZ07, k is 15, the value of probIdx is equal to (P>>8), rangeIdx is equal to ((Range>>6)&3). If P is equal to or greater than 16384 (or if the 15th order bit of P is equal to 1), RangeOne is equal to (range - RangeSH) and RangeZero is equal to RangeSH. Otherwise (that is, if P is less than 16384), RangeOne is equal to RangeSH and RangeZero is equal to (range - RangeSH). Table 17

[0067] In another example, the 8-column reflected RangeSH table is used as shown in Table 18. For JCTVC-F254 and VCEG-AZ07, the parameter settings correspond to k = 15, n = 7, and m = 3. The related probability parameters are derived as probIdx = (P >>8), rangeIdx = ((range>>5) &7). If P is equal to or greater than 16384 (that is, if the 15th order bit of P is 1), the related probability parameters are derived as RangeOne = (range - RangeSH) and RangeZero = RangeSH. Otherwise (that is, if P is less than 16384), the related probability parameters are derived in the form of RangeOne = RangeSH and RangeZero = (range - RangeSH). Table 18

[0068] In another example, the 8 column by 64 row reflected RangeSH table is used as shown in Table 19. For JCTVC-F254 and VCEG-AZ07, the parameter settings correspond to k = 15, n = 6, in = 3. Related probability parameters are derived as probIdx = (P >>9), rangeIdx = ((range>>5) &7). If P is equal to or greater than 16384 (that is, if the 15th order bit of P is 1), the related probability parameters are derived as RangeOne = (range - RangeSH) and RangeZero = RangeSH. Otherwise (that is, if P is less than 16384), RangeOne = RangeSH and RangeZero = (range - RangeSH). The table size of Table 19 is the same as the HEVC or H.264/AVC RangeSH table. Table 19

[0069] In Tables 17 to 19, the RangeSH launch value will not necessarily be trimmed to be greater than 3.

[0070] Table 20 illustrates the comparison of the range lookup table size of an embodiment of the present invention and the JCTVC-F254 with the HEVC standard. The JCTVC-F254 based approach requires 12126% of the HEVC record size whereas the embodiment of the present invention requires 870% of the HEVC record size. In other words, the method described in CTVC-F254 requires a look-up table practically 14 times larger than that of the embodiment of the present invention. Table 20

[0071] In another embodiment, if the probability of one is greater than 0.5 (0.64, for example), the probability greater than 0.5 will be used to query the table. If the range is 500, for example, the fourth column is used. The probability value of 0.14 corresponds to the value in the 47th row, which is 68. RangeOne can be ((Range Mid/2) + 68) = 308 or it can be ((range/2 + 58) = 318. If RangeOne is greater than Range, RangeOne can be trimmed to (range - K), where K is an integer and K can be different for different range values or for different sections.

[0072] In the priority-based filter block fusion scheme, the first step consists of choosing the maximum number N of candidates from M predefined neighboring blocks. However, the number of available filter candidates among the M predefined neighboring blocks may be less than N due to unavailability at image boundaries, repetitive filters, or out-filtering. When this case occurs, some loss of encoding performance may occur. To overcome this problem, the following embodiments are described, in which one or more filters are added to the candidate list of the priority-based filter block fusion scheme. In the first embodiment, additional filters are generated using available filters. Some coefficients, for example, far from the center position can be removed to form a new filter. In another example, a new symmetric filter can be generated by averaging the coefficients of an available filter. In yet another example, one or more predefined filters, or a filter that is the average of available filters, may be added to populate the candidate list of the priority-based filter block fusion scheme.

[0073] In the embodiments described above, the similarity between the pixel to be filtered and neighboring pixels can be used for pixel classification. Neighboring pixels are defined using a window such as a cross pattern, a 3x3 square, or a 5x5 diamond. The central position of a window is that of the pixels to be filtered. For each neighboring pixel in this window, if the pixel value difference between the neighboring pixel and the pixel to be filtered is less than a threshold, the similarity is increased by one. Otherwise, the similarity is not increased. After comparing all neighboring pixels to the pixel to be filtered, we can obtain a similarity value for a pixel to be filtered. Based on the similarity values, pixels can be classified into different groups, and different filters are applied to different groups in the ALF (adaptive circuit filter) process. During the development of the HEVC standard, the original pixel classification was applied to all pixels in an image in accordance with HM-7, or before HM-7.0. However, other adaptive schemes such as the CTB-based ALF scheme (based on coding tree blocks) were also described during the development of the HEVC standard, in which ALF parameters are encoded and can be changed from CTB to CTB. It is also possible to apply a pixel classification to some CTBs in an image.

[0074] Figure 5 illustrates an exemplary flowchart for another context-based adaptive binary arithmetic coding (CABAC) based on multiple tables in accordance with an embodiment of the present invention. The method encodes or decodes a current bin of a set of binary data from a current encoding symbol according to a probability of a binary value of the current bin as shown in step 510, wherein the probability of the binary value of the current bin is generated from one or more previously encoded symbols before the current encoding symbol. The probability of the binary value is updated according to the binary value of the current bin for the next bin using multi-parameter probability models in step 520, where each of the multi-parameter probability models is updated using an individual set of probability states associated with a corresponding parameter.

[0075] Figure 6 illustrates an exemplary flowchart for another context-based adaptive binary arithmetic coding (CABAC) based on multiple tables, in accordance with an embodiment of the present invention. The method encodes or decodes a current bin of a set of binary data from a current encoding symbol according to a probability of a binary value of the current bin as shown in step 610, wherein the probability of the binary value of the current bin is generated from one or more previously encoded symbols before the current encoding symbol. The probability of the binary value is updated according to the binary value of the current bin for a next bin in step 620, where RangeOne or RangeZero values derived from at least one range lookup table are used for encoding or decoding. of the current bin, and in which a smaller range half (RangeSH), RangeOne or RangeZero values are derived for a given range interval based on an average range value of the given range interval and a given probability of the binary value.

[0076] The flowcharts shown are intended to illustrate an example of video coding in accordance with the present invention. Those skilled in the art may modify each step, rearrange steps, divide a step or combine steps to implement the present invention without deviating from the spirit of the present invention. In the description specific syntax and semantics were used to illustrate examples for implementing embodiments of the present invention. Those skilled in the art will be able to practice the present invention by replacing the syntax and semantics with equivalent syntax and semantics without deviating from the spirit of the present invention.

[0077] The above description is presented to enable those skilled in the art to put into practice the present invention as proposed in the context of a specific application and its requirements. Various modifications to the described embodiments will be apparent to those skilled in the art, and the general principles defined herein may be applied to other embodiments. Therefore, the present invention is not intended to be limited to the specific embodiments shown and described, but is given the broadest scope consistent with the principles and novel features disclosed herein. In the above detailed description, several specific details are illustrated to provide a comprehensive understanding of the present invention. However, it should be understood by those skilled in the art that the present invention can be implemented.

[0078] The embodiment of the present invention as described above can be implemented in various hardware, software codes or by a combination of the two. An embodiment of the present invention, for example, may consist of one or more circuits integrated into a video compression chip or program code integrated into video compression software to perform the processing described herein. An embodiment of the present invention may also consist of program code to be executed on a Digital Signal Processor (DSP) to conduct the processing described herein. The invention may also involve a series of functions to be driven by a computer processor, a digital signal processor, a microprocessor or a field programmable gate array (FPGA). These processors can be configured to perform specific tasks in accordance with the invention by executing software code or machine-readable firmware code that defines the specific methods embodied by the invention. Software code or firmware code can be developed in different programming languages and in different formats or styles. The software code can also be compiled for different target platforms. However, different code formats, software code styles and languages, and other means of configuring codes to perform tasks according to the invention will not deviate from the spirit and scope of the invention.

[0079] The present invention can be incorporated in other specific forms without deviating from its spirit or its essential characteristics. The examples described should be considered in all respects as illustrative and not restrictive. The scope of the present invention is therefore indicated by the appended claims and not by the above description. All changes that affect the meaning and scope of equivalence of the claims must be included in their scope.

Claims (8)

1. Method of entropy coding image or video data in an image or video encoder or decoder, the method CHARACTERIZED by the fact that it comprises: encoding or decoding a current bin of a binary data set of a symbol current encoding according to a probability of a binary value of the current bin, wherein the probability of the binary value of the current bin is generated from one or more symbols previously encoded before the current encoding symbol; and updating the probability of the binary value according to the binary value of the current bin for a next bin, wherein the range One or range Zero values derived from at least one range lookup table are used for encoding or encoding. decoding the current bin; and wherein a lower half of the range (RangeSH), range One or range Zero values are derived for a given range range based on the given range range and a given probability of the binary value. 2. Method, according to claim 1, CHARACTERIZED by the fact that at least one range lookup table comprises a zero range table or a one range table. 3. Method according to claim 2, CHARACTERIZED by the fact that a range value is derived for an average range value of the given range interval and an average probability value of the given range interval from a Zero probability range or probability range One. 4. Method according to claim 3, CHARACTERIZED by the fact that a range value is derived for an average range value of the given range range and a maximum probability value of the given range range from a Zero probability range or probability range One. 5. Method according to claim 2, CHARACTERIZED by the fact that at least one range lookup table only includes range values for a Zero probability range or a One probability range between 0.0 and 0.5. 6. Method, according to claim 5, CHARACTERIZED by the fact that the range value of the probability range Zero or probability range One between 0.5 and 1.0 are derived by: current range - range values for range probability One or probability Zero range between 0.0 and 0.5. 7. Method according to claim 2, CHARACTERIZED by the fact that at least one range lookup table includes range values for probability range Zero or probability range One between 0.0 and 1.0, and wherein the range values for the Zero probability range or the One probability range between 0.5 and 1.0 are reflected from the range values for the Zero probability range or the One probability range between 0.0 and 0.5 . 8. Entropy coding apparatus for an image or video encoder or decoder, the entropy coding apparatus CHARACTERIZED by the fact that it comprises: a binary arithmetic encoder for encoding or decoding a bin binary data stream of a symbol current encoding according to a probability of a binary value for the current bin, wherein the probability of the binary value of the current bin is generated from one or more symbols previously encoded before the current encoding symbol; and a context model unit for updating the probability of the binary value according to the binary value of the current bin for a following bin, wherein the range One or range Zero values derived from at least one range lookup table are used to encode or decode the current bin; and wherein the range One or range Zero values are derived for a given range interval based on an average range value of the given range interval.