WO2019191886A1

WO2019191886A1 - Probability update method and apparatus for context

Info

Publication number: WO2019191886A1
Application number: PCT/CN2018/081648
Authority: WO
Inventors: 马思伟; 崔静; 王苫社; 郑萧桢
Original assignee: 北京大学; 深圳市大疆创新科技有限公司
Priority date: 2018-04-02
Filing date: 2018-04-02
Publication date: 2019-10-10
Also published as: CN110495170A

Abstract

Provided are a probability update method and apparatus for context. The method comprises: when the number of times a binary symbol is coded/decoded using a probability update model for context meets a condition, adjusting the probability update model; and based on the adjusted probability update model, coding/decoding the binary symbol. According to the number of times that coding/decoding is carried out using a probability update model for context, the probability update model is adjusted, such that the coding/decoding performance can be improved to some extent.

Description

Context probability update method and device

Copyright statement

The disclosure of this patent document contains material that is subject to copyright protection. This copyright is the property of the copyright holder. The copyright owner has no objection to the reproduction of the patent document or the patent disclosure in the official records and files of the Patent and Trademark Office.

Technical field

The present application relates to the field of video codec and, more particularly, to a method and apparatus for probabilistic updating of a context.

Background technique

Context-based Adaptive Binary Arithmetic Coding (CABAC) is an entropy coding scheme adopted by the international video coding standard H.264. The CABAC method is also adopted in the next-generation HEVC standard of H.264 and the codec standards currently under development.

The syntax elements in the entropy coding are binarized to obtain a plurality of binary bits, each of which is called a bin, and the binary bit string formed by the multiple binary bits is called a bin string. Each bin or a number of bins in a bin string is assigned a context, and the model-related parameters are updated by the probability stored in the context, and the bin string is encoded by CABAC. The general process of encoding is to predict the probability value of the bin using a probability update model, and then arithmetically encode the bin according to the probability value.

At present, the coding performance of CABAC needs to be further improved and optimized.

Summary of the invention

The present application provides a context probability update method and apparatus, which can further improve the encoding/decoding performance of CABAC.

In a first aspect, a context update method is provided, the method comprising: adjusting a probability update model when a number of binary symbols for encoding/decoding binary symbols is satisfied using a probability update model of the context Binding/decoding binary symbols based on the adjusted probability update model.

In the solution provided by the present application, the probability update model is adjusted according to the number of times of encoding/decoding of the model using the probability of using the context, so that the encoding/decoding performance can be improved to some extent.

In a second aspect, a context probability updating apparatus is provided, the apparatus comprising: a processing unit, configured to: when a number of binary symbols for encoding/decoding a binary symbol is satisfied using a probability update model of the context, The probability update model is adjusted; an encoding/decoding unit is configured to encode/decode the binary symbols based on the adjusted probability update model.

In a third aspect, an encoder is provided, the encoder comprising a memory and a processor, the memory for storing instructions, the processor for executing the instructions stored by the memory, and instructions stored in the memory Execution of the processor causes the processor to perform the method of encoding provided by the first aspect.

In a fourth aspect, a decoder is provided, the decoder comprising a memory and a processor, the memory for storing instructions, the processor for executing the instructions stored by the memory, and instructions stored in the memory Execution of the processor causes the processor to perform the method of decoding provided by the first aspect.

In a fifth aspect, a chip is provided, the chip comprising a processing module and a communication interface, the processing module is configured to control the communication interface to communicate with an external, and the processing module is further configured to implement the method provided by the first aspect.

In a sixth aspect, a computer readable storage medium is provided having stored thereon a computer program that, when executed by a computer, causes the computer to implement the method provided by the first aspect. In particular, the computer is for example an encoder/decoder.

In a seventh aspect, a computer program product comprising instructions, which when executed by a computer, causes the computer to implement the method provided by the first aspect. In particular, the computer is for example an encoder/decoder.

Therefore, in the solution provided by the present application, the probability update model is adjusted according to the number of times of encoding/decoding of the model using the probability of using the context, so that the encoding/decoding performance can be improved to some extent.

DRAWINGS

FIG. 1 is a schematic flowchart of a method for updating a probability of a context provided by an embodiment of the present application.

FIG. 2 is another schematic flowchart of a method for updating a probability of a context provided by an embodiment of the present application.

FIG. 3 is a schematic block diagram of a probability update apparatus for a context provided by an embodiment of the present application.

FIG. 4 is a schematic block diagram of an encoder according to an embodiment of the present application.

FIG. 5 is a schematic block diagram of a decoder provided by an embodiment of the present application.

detailed description

In order to facilitate understanding and description of the solutions provided by the embodiments of the present application, the following concepts are first described below.

1. CABAC encoding process.

1) Binarization.

The syntax elements in entropy coding are converted to Binarization. After converting the syntax elements into binary, a plurality of binary symbols are obtained, each of which is called a bin, and the binary bit string formed by the plurality of binary symbols is called a bin string.

If the symbol to be encoded is already a binary symbol, step 1) can be omitted.

2) Select the probability update model.

Each bin or bins in a bin string is assigned a context, and the model-related parameters are updated by the probability stored in the context, and the bin string is encoded using CABAC. For a bin to be encoded, select the probability update model stored in the context to which it is assigned.

3) Binary arithmetic coding.

The binary arithmetic coding of the bin is performed by the selected probability update model.

The decoding process of CABAC corresponds to the coding process of CABAC.

2. Probability update model.

In the process of arithmetic coding, the probability contained in the same context or the information contained in the indicated probability is not fixed, but is updated as the number of bins that decode the context is changed. Specifically, the probability included in the context or the information included to indicate the probability is updated based on the probability update model of the context.

In different implementations, the number of probability update models may be different, and the number of probability update models is not limited in this application.

As an implementation, a probability update model is used to update the probability in the context or the information pointing to the probability.

When bin is "1", U[k+1]=U[k]+((2^15-U[k])>>4),

When bin is "0", U[k+1]=U[k]-(U[k]>>4),

P[k+1]=U[k+1] formula (1)

Where P[k+1] is the final prediction probability and U is the probability update model. k and k+1 represent the current state and the next state, respectively. 4 is a sliding window parameter.

As another implementation, the probability of predicting bin is estimated using two probability update models.

When bin is "1",

U[k+1]=U[k]+((2^15-U[k])>>4),V[k+1]=V[k]+((2^15-V[k] )>>8),

When bin is "0",

U[k+1]=U[k]-(U[k]>>4), V[k+1]=V[k]-(V[k]>>8),

P[k+1]=(U[k+1]+V[k+1])>>1 formula (2)

Among them, P[k+1] is the final prediction probability, and U and V are two probability update models. k and k+1 represent the current state and the next state, respectively. 4 is the sliding window parameter in the probability update model U, and 8 is the sliding window parameter in the probability update model V.

It should be noted that the expression of the probability update model may be different in different standards, because the probability update model is a process of maintaining the probability value in the arithmetic coding, and the manner in which the process is implemented in different standards is different. For example, in some video standards, the probabilistic update model may implement an update of the model in conjunction with a state jump of a pre-designed finite state machine that specifies the current state transition to the next state rule. The state value of the finite state machine expresses probability related information.

In some video standards, it is preferred to use a context update model to express a probability update model, and accordingly, to use contextual modeling instead of a probabilistic update process.

In some video standards, the most important thing stored in the context model is the information related to the probability update. The related information expresses an implicit value of the probability value, such as the state value of the finite state machine described above. Alternatively, the probability related information is an explicit form of probability correlation value, such as a probability value. Regardless of whether the context model stores a probability-dependent value in the form of a display or an implicit form, the modeling of the context specifies what rules the probability of the context is updated with.

In the current practice, the sliding window parameter takes a fixed value, that is, the probability update model of the context remains unchanged.

3. The initialization cycle of the context.

In the entropy coding process of CABAC, the context is initialized with a certain period, which is called the initialization period of the context. For example, the context is initialized with a period of a slice, that is, the initialization period of the context is a slice.

At the same time as the context is initialized, the context's probability update model is also initialized. Initializing the probability update model includes initializing the sliding window parameters in the probability update model.

For example, in the implementation described above using a probability update model prediction probability, the value of the initialized sliding window parameter is 4. For another example, in the implementation described above using two probability update model prediction probabilities, the values of the two sliding window parameters after initialization are 4 and 8, respectively.

In the prior art, the probability update model remains unchanged after initialization until initialization is initiated again at the beginning of the next initial cycle.

The present application provides a method and apparatus for probabilistic updating in CABAC, which can further improve encoding (or decoding) performance relative to the prior art.

The solution provided by the present application can be applied to coding as well as to decoding.

FIG. 1 is a schematic flowchart of a method for updating a probability of a context provided by an embodiment of the present application. This method can be performed by an encoder/decoder. The method includes the following steps.

S110: When the number of times the binary symbol is encoded/decoded by using the probability update model of the context satisfies the condition, the probability update model is adjusted.

The number of times the binary symbol is encoded/decoded using the probability update model of the context satisfies the condition that the number of times the binary symbol is encoded using the probability update model of the context satisfies the condition, or the probability of the context is updated using the context to update the model. The number of decodings satisfies the condition.

In some embodiments, the number of times the binary symbol is encoded/decoded using the probability update model of the context satisfies the condition, meaning that the number of times the binary symbol is encoded/decoded using the probability update model of the context reaches a threshold.

The number of times the binary symbol is encoded/decoded using the probability update model of the context reaches a threshold, which means that the number of times the binary symbol is encoded using the probability update model of the context reaches a threshold, or the binary symbol is updated using the probability update model of the context. The number of decodings reaches the threshold.

Specifically, in S110, when the number of times the binary symbol is encoded/decoded by the probability update model of the context reaches a threshold, the probability update model is adjusted.

In the coding scenario, when the number of bins using the probability update model code reaches the threshold, the probability update model is adjusted.

In the decoding scenario, when the number of decoding bins using the probability update model reaches the threshold, the probability update model is adjusted.

The method of determining the threshold will be described below.

In some embodiments, the number of times the binary symbol is encoded/decoded using the probability update model of the context satisfies the condition, and may also represent other situations, for example, the number of times the binary symbol is encoded/decoded using the probability update model of the context. The duration exceeds the preset value. This application does not limit this, as long as the number of encoding/decoding binary symbols is obtained by using the probability update model of the context to obtain a better timing for adjusting the probability update model.

Adjusting the probability update model refers to changing the probability update model so that the adjusted and updated probabilistic update model is different.

Optionally, adjusting the probability update model includes adjusting a value of one or more parameters in the probability update model.

For example, adjust the probability of updating the sliding window parameters in the model. For another example, adjust the probability to update other parameters in the model. This application does not limit this.

Optionally, adjusting the value of the one or more parameters in the probability update model includes: reducing or increasing the value of the parameter in the probability update model.

S120: The binary symbol is encoded/decoded by using the adjusted probability update model.

In the coding scenario, in S110, when the number of times the binary symbol is encoded using the probability update model of the context satisfies the condition, the probability update model is adjusted, and in S120, the probability of the model is predicted using the adjusted probability update model. And using the probability, binary arithmetic coding of the bin.

In the decoding scenario, in S110, when the number of times the binary symbol is decoded using the probability update model of the context satisfies the condition, the probability update model is adjusted, and in S120, the probability of the model is predicted using the adjusted probability update model. And using the probability, binary arithmetic decoding of the bin.

In the coding scenario, in the process of using the probability update model of the context to perform arithmetic coding, when the coding is started, the already coded information can be referenced less, and as the coding proceeds, the already coded information can be referred to more and more. Many, if you always use a fixed probability to update the model for encoding, it may not be conducive to the improvement of encoding performance.

In the decoding scenario, in the process of performing arithmetic decoding by using the probability update model of the context, when the decoding is started, the already decoded information can be referred to less, and as the decoding progresses, the already decoded information can be referred to more and more. Many, if you always use a fixed probability to update the model for decoding, it may not be conducive to the improvement of decoding performance.

In the embodiment of the present application, the probability update model is adjusted according to the number of times of encoding/decoding of the model using the probability of using the context. For example, when the number of times of encoding/decoding of the probability update model using the context reaches the threshold, the probability update model is adjusted, and it can be considered that there is already a lot of already encoded information that can be referred to at this time, and the probability update model needs to be adjusted to adapt to the change. The encoding environment, which can improve the encoding/decoding performance to some extent.

It should be understood that the decoding process corresponds to the encoding process, and the technical solution provided by the present application can be applied to the encoding process as well as to the decoding process. For convenience of description, some embodiments herein may only describe an encoding scenario (or a decoding scenario) as an example, and related schemes and technical effects thereof are equally applicable to a decoding scenario (or a coding scenario).

Optionally, in S110, when the number of times the binary symbol is encoded/decoded by the probability update model of the context reaches the threshold in an initialization period, the probability update model is adjusted; in 120, in the same The binary symbols are encoded/decoded using the adjusted probability update model during the initialization cycle.

Taking the coding scenario as an example, at the beginning of an initialization period, the probability update model of the context is initialized; after the initialization is completed, the probability of the model prediction bin is updated using the probability, and the binary probability encoding is performed on the bin using the predicted probability; When the number of bins encoded by the probability update model reaches the threshold, the probability update model is adjusted; in the initialization period, the adjusted probability is used to update the model to continue predicting the probability of bin, and the binary probability is used to perform binary arithmetic on the bin. coding.

Taking the decoding scene as an example, at the beginning of an initialization period, the probability update model of the context is initialized; after the initialization is completed, the probability of the model prediction bin is updated using the probability, and the binary probability decoding of the bin is performed using the predicted probability; When the number of bins to be decoded by the probability update model reaches the threshold, the probability update model is adjusted; in the initialization period, the adjusted probability is used to update the model to continue predicting the probability of the bin, and the predicted probability is used to perform binary arithmetic on the bin. decoding.

In the prior art, after the initialization of the probability update model, the probability update model is fixed during the current initialization period, and in an initialization period, as the encoding (or decoding) proceeds, the referenced encoding may be performed (or More and more information is being decoded, and updating the model with a fixed probability may be detrimental to the performance of the encoding (or encoding).

In the embodiment of the present application, the probability update model is adjusted according to the probability of updating the model according to the probability of using the context in the initialization period of the same context, and the encoding/decoding performance can be improved to a certain extent.

A frame of image consists of one or more slices, and a slice can be split into one or more coding tree units (CTUs).

Optionally, in some embodiments, the initialization of the context may be in a period of one slice, that is, the initialization period of the context is one slice.

Optionally, in some embodiments, the initialization of the context may be in a CTU period, that is, the initialization period of the context is one CTU.

Optionally, in other embodiments, the initialization of the context may also be cycled with one other image coding unit. That is, the initialization period of the context may be other image coding units, which is not limited in this application.

In some embodiments, prior to encoding/decoding the binary symbols using the probability update model of the context, the method further comprises: initializing the probability update model.

Specifically, the parameters in the probability update model are initialized. For example, the sliding window parameters in the probability update model are initialized.

In some embodiments, the sliding window parameters may be initialized according to a quantization parameter (QP) of the current slice.

In some embodiments, the sliding window parameters may be initialized based on the quantization step size of the current CTU.

In some embodiments, the sliding window parameters of the probability update model are initialized based on the quantization step size of the block currently being encoded or decoded.

In some embodiments, the sliding window parameters may also be initialized based on the encoding state of the current context.

In the present application, when the number of times the binary symbol is encoded/decoded by the probability update model of the context reaches the threshold, the probability update model is adjusted, wherein the threshold may be determined by the following implementation manners.

In the first implementation, the threshold is determined based on the probability state value of the arithmetically encoded zoom interval two-dimensional lookup table.

In some embodiments, the threshold is a positive integer that is less than a maximum of the probability state value.

Assuming that the maximum value of the probability state value of the two-dimensional lookup table is 512, the threshold is a positive integer less than 512, for example, the threshold is 128.

The arithmetic interval of the arithmetic interval is a coding table in CABAC. As the name implies, the arithmetic interval of the scaled two-dimensional lookup table is a two-dimensional table. The scaling interval two-dimensional lookup table of the arithmetic coding may also be simply referred to as an interval query table. For the sake of brevity, the description of the interval lookup table is used below.

The interval query table is used to query the current probability and the size of the new scaling interval corresponding to the scaling interval of the arithmetic coding engine. The interval query table is an implementation of the CABAC-free multiplication method, that is, avoiding the multiplication operation between the probability and the size of the current zoom interval to obtain the size of the new zoom interval.

The interval lookup table has two index values: a probability state value and a scaling interval value. The scale interval value represents the length of the interval in the interval lookup table (Range). The probability state value represents the probability value of each interval.

These two index values, the probability state value and the scaling interval value, are used to query the new zoom interval of CABAC.

In some implementations, the symbol "x" is used to represent the probability state value, and the symbol "y" is used to represent the current scaling interval value of the arithmetic coding engine. For simplicity of description, the symbol "x" is used hereinafter to represent the probability state value, and the symbol "y" is used to represent the current scaling interval value of the arithmetic coding engine. However, this application does not limit this. In other implementations, other symbols may also be used to represent the probability state value and the current scaling interval value of the arithmetic coding engine.

The probability state value x represents a probability value, that is, a value of P[k+1] in the example of the probability update model described above). Assuming that the range of U and V is [1, 2^15], the interval of P[k+1] is the same. However, the value of x is not 2^15, because the table is too large, so in some embodiments, x is P[k]>>6 (that is, 2^9=512). In some embodiments, the value of x may also be 1024 or greater.

The value of the scale interval value y represents the interval length (Range), and Range ranges from 1 to 512. In some embodiments, y = (Range>> 2) - 64. The reason for this is that Range can be regarded as a 9-bit binary, and shifting two bits to the right means that only the size of the upper 7 bits is considered, so the value of the scaling interval value y here is the high 7 bits minus 64. Regardless of the operation, the goal is to quantify the interval. The range of the final scaling interval value y is [0, 64].

For example, the Range value is between 488 (111101000) and 491 (111101011), that is, the value of Range is shifted to two by 122 (1111010), and after subtracting 64, it is 58, then the value of Range is in the range of 488-491. Its corresponding index value is 58. For another example, the Range value is in the range of 492 (111101100) to 495 (111101111), the value of Range is shifted to two by 123 (1111011), and after subtracting 64, it is 59, then the Range value is in the range of 492-495. Its corresponding index value is 58.

The above method of determining the values of the probability state value x and the scaling interval value y is merely an example and not a limitation. In practice, it can be implemented in other ways. For example, the value of Range can be directly used without determining the value of the scaling interval value y.

In some implementations, the size of the interval lookup table is 512 x 64. That is, x ranges from 1 to 512, and y ranges from 1 to 64.

In other implementations, the maximum value of x and y may be 1024 or greater.

For example, when the size of the interval lookup table is 512 x 64, the threshold may be set to a positive integer less than 512.

In some embodiments, the threshold is an integer between X/2 ⁿ¹ and X/2 ⁿ² , where X is the maximum value of the probability state value of the arithmetically encoded zoom interval two-dimensional look-up table, X/2 ^{n1 is} greater than 1 , n1 and n2 are natural numbers, and n1 is greater than n2.

For example, the maximum value X of the probability state value of the scaled interval two-dimensional lookup table of the arithmetic coding is 512, and the threshold may be an integer between 512/2 ⁴ and 512/2 ² (ie an integer between 32 and 128), Or the threshold may be an integer between 512/2 ³ and X/2 ¹ (ie an integer between 61 and 256).

The above examples are merely examples and are not limiting. In some standards, the maximum value X of the probability state values of the scaled two-dimensional look-up table of the arithmetic coding may also be other values, such as 1024 or the like.

In a second implementation, the threshold is adaptively adjusted.

The threshold is determined based on the average number of uses of the probability update model of the same context over one or more initialization cycles that have been encoded/decoded.

For example, the average number of encoding/decoding times of the current context in one or more initialization periods that have been encoded/decoded is counted, and then the value of the threshold is determined according to the average number of encoding/decoding.

In the coding scenario, the probability update model of one context is encoded in the already-initialized initialization period T1, and the number of times of encoding in the already-initialized initialization period T2 is N2, which is performed in the already-encoded initialization period T3. The number of times of encoding is N3, and N1, N2, and N3 are averaged, and the average value obtained is the average number of times the probability update model of the context is used in the three initialization periods.

In the decoding scenario, the number of times the probability update model of one context is decoded in the already decoded initialization period T1 is N1, and the number of times of decoding in the already decoded initialization period T2 is N2, and is performed in the already initialized initialization period T3. The number of decodings is N3, and N1, N2, and N3 are averaged, and the average value obtained is the average number of times the probability update model of the context is used in the three initialization periods.

As another example, the average number of encodings of a plurality of different contexts within one or more initialization periods that have been encoded/decoded is counted, and then the value of the threshold is determined based on the average number of encodings.

Specifically, the average number of encodings of the context within one or more initialization periods that have been encoded/decoded may be counted by pre-training.

When the calculated average usage count is less than the first value, the threshold is set to the first value.

When the calculated average usage count is greater than the second value, the threshold is set to the second value, the first value being less than the second value.

When the calculated average usage count is greater than or equal to the first value and less than or equal to the second value, the threshold is set to be greater than the first value and less than the integer of the second value.

The first value is Thrmin, the second value is Thrmax, and the average number of uses is Numave. The threshold Thr is determined according to the following formula:

Wherein, the first value and the second value respectively represent the minimum value and the maximum value of the threshold. The first data and the second value can be determined based on empirical values. For example, the first value is 32 and the second value is 512.

In this embodiment, the value of the threshold is adaptively changed according to the average usage number of the probability update model of the same context in one or more initialization periods that have been encoded/decoded, so that it can be flexibly and appropriately The timing adjusts the probability update model, so the coding performance can be further improved.

In a third implementation, the threshold is determined according to an initialization period of the context.

For example, when the initialization period of the context is one slice, the threshold is set to 512. When the initialization period of the context is one CTU, the threshold is set to an integer between 32 and 512.

Three methods of determining the threshold are described above. The present application is not limited thereto. In some implementations, the threshold may also be determined according to other feasible methods, as long as the probability update model can be adjusted at an appropriate timing.

In some embodiments, the number of times a binary symbol is encoded/decoded using a probability update model can be counted by a counter.

When the probability update model is initialized, the counter of each context is also initialized. When the binary symbol is encoded/decoded using the probability update model, the counter also starts counting, and each time the encoding/decoding is performed using the context, the corresponding The counter is counted once, and when the count value of the counter reaches the threshold, it is determined that the number of binary symbols for encoding/decoding binary symbols using the probability update model reaches the threshold.

For a context, the context includes a counter that counts the number of encoding/decoding binary symbols used to update the model using the probability of the context. In other words, each context has its own counter, and when a bin references a context, the context counter is incremented by one.

In some embodiments, the counter is cleared when the current stripe encoding/decoding ends.

In some embodiments, the counter is cleared when the current coding tree unit is encoded/decoded.

In some embodiments, the counter is cleared at the end of the current initialization period.

In some embodiments, the counter is cleared when the counter value of the counter reaches the threshold.

For example, if it is necessary to make multiple adjustments to the probability update model in an initialization cycle, in this case, when the count value of the counter reaches the threshold, the counter is cleared, and restarts with the next encoding/decoding start. count.

In some embodiments, the counter is used to update the model to encode/decode the number of binary symbols, and when the counter's count reaches the threshold 1, the probabilistic update model is first adjusted while the counter is cleared; The counter count uses the probability update model to encode/decode the number of binary symbols. When the counter count reaches the threshold 2, the probability update model is adjusted a second time and the counter is cleared. The threshold 1 and the threshold 2 may be the same or different. The manner of the first adjustment and the second adjustment of the probability update model may be the same or different. This embodiment is described by taking two adjustments to the probability update model as an example. However, the present application is not limited thereto. In actual operation, the probability update model can be adjusted more times, the process of multiple adjustments, and two adjustments. The process is similar and will not be repeated here.

In some embodiments, when the number of binary symbols encoded/decoded with the probability update model does not reach the threshold, but the current initialization period ends, the counter is also cleared.

Optionally, in some embodiments, adjusting the probability update model in S110 specifically includes: adjusting a sliding window parameter in the probability update model.

The sliding window parameter can determine the speed of the probability update of the probability update model. The larger the sliding window parameter, the smaller the probability update speed of the probability update model, and vice versa. Therefore, since the sliding window parameter can determine the speed of the probability update of the probability update model, by adjusting the sliding window parameter, the probability that the probability update model can be adjusted for the probability update can be adjusted.

Taking the coding scenario as an example, when coding is first started, there is less information that can be referenced, and the speed can be updated with a faster probability; as the coding proceeds, more and more information can be referenced. At this time, you should use a slower probability to update the speed. It should be understood that as the number of encodings increases, the speed of the probability update is adjusted, and the probability update speed and probability update accuracy of the context can be better balanced.

In the prior art, after initializing the probability update model, the sliding window parameters are always maintained.

In the embodiment of the present application, when the number of times the binary symbol is encoded/decoded reaches a threshold by using the probability update model, the sliding window parameter in the probability update model is adjusted, thereby adjusting the probability update probability of the probability update model, and thus balancing The probability update speed and probability update accuracy, which can further improve the coding performance.

In addition, in this embodiment, the size of the sliding window parameter in the probability update model is adjusted by updating the model according to the number of times the binary symbol is encoded/decoded according to the usage probability, so that the number of times the binary symbol is encoded/decoded using the probability update model and the sliding The values of the window parameters are related, which can improve the accuracy of the probability prediction of the context.

Optionally, in some embodiments, adjusting the probability update model in S110 may further include: adjusting other parameters in the probability update model. For example, other parameters may affect the speed and/or accuracy of the probability update of the probability update model.

In some embodiments, each context has its own probability update model, ie each context has its own sliding window parameters. The sliding window parameters in a context's probability update model are stored in that context.

Optionally, in some embodiments, as shown in FIG. 2, in S110, when the number of times the binary symbol is encoded/decoded using the probability update model of the context reaches a threshold, the sliding window parameter is increased.

When the sliding window parameter is increased, the probability of the probabilistic update of the probability update model can be reduced.

Taking the coding scenario as an example, when using the probability update model for arithmetic coding, at the beginning, there is less information that can be referenced. At this time, a faster probability update speed can be used; as the coding proceeds, reference can already be made. More and more information is encoded, and a slower probability update speed should be used. In this embodiment, when the number of times of encoding using the probability update model reaches the threshold, the speed of the probability update is reduced by increasing the sliding window parameter, so that the encoding performance can be improved.

Taking the decoding scene as an example, when using the probability update model for arithmetic decoding, at the beginning, the already decoded information can be referenced less. At this time, the faster probability update speed can be adopted; as the decoding progresses, the reference can already be used. More and more information is decoded, and a slower probability update speed should be used. In this embodiment, when the number of times of decoding using the probability update model reaches the threshold, the speed of the probability update is reduced by increasing the sliding window parameter, so that the decoding performance can be improved.

Optionally, in some embodiments, adjusting the sliding window parameter in the probability update model in S110, specifically: reducing the sliding window parameter.

For example, due to actual demand, if the number of times the binary symbol is encoded/decoded by using the probability update model reaches the threshold, the speed of the probability update needs to be accelerated, and at this time, the binary symbol can be encoded/decoded using the probability update model. When the number of times reaches this threshold, the sliding window parameters are reduced.

As described above, in the embodiment of the present application, after the probability update model is initialized, the sliding window parameter may be adjusted rather than fixed, so that the speed of the probability update may be adjusted, thereby balancing the probability update speed and the probability update accuracy, thereby Can improve coding performance.

Optionally, in some embodiments, bins are encoded/decoded using a plurality of probabilistic update models, each probabilistic update model having a sliding window parameter, in this case having a plurality of sliding window parameters, S110 specifically including : When the number of times the binary symbol is encoded/decoded using the plurality of probability update models reaches the threshold, the plurality of sliding window parameters are increased.

In some example manners, the magnitude of the increase in the different sliding windows may be the same or different.

In an embodiment in which the sliding window parameter is increased, if the probability state value of the arithmetically encoded zoom interval two-dimensional lookup table is less than the third value, the sliding window parameter is increased by a magnitude less than the fourth value. Wherein, the third value and the fourth value may be empirical values.

For example, the third value can be 11 and the fourth value can be 2.

The reason for this is that when the probability state value x of the arithmetically encoded zoom interval two-dimensional lookup table is smaller than the third value, the probability value of the context is already very small, and if a large sliding window parameter is used for the probability update, The probability of several bins using this context that result in consecutive encodings is very close, which is of little significance for arithmetic coding. Therefore, when the probability state value x of the arithmetically encoded zoom interval two-dimensional lookup table is smaller than the third value, the sliding window parameter should be increased in amplitude to be small.

In some embodiments, during an initialization period, the probability update model needs to be adjusted multiple times, that is, the sliding window parameters are adjusted multiple times, wherein the amplitude values of each adjustment may be the same or different.

In some embodiments, the sliding window parameter in the probability update model has a value range of [Min, Max]. That is, the value of the sliding window parameter before or after the adjustment is greater than or equal to Min and less than or equal to Max.

In some embodiments, Min is 2 and Max is 9.

Let's take a specific probability update model as an example to explain briefly why the sliding window parameter has a minimum value of 2 and a maximum value of 9.

Take the probability update model shown in the following formula (3) as an example:

Where P is the probability update model, a is the sliding window parameter, k and k+1 are the current device and the next state, respectively, and α is a value located in the interval (0, 1), for example

In the probability update model shown in equation (3), it is assumed that the sliding window parameter a is 2, that is, the value of α is

Then, in the case where bin is "0", P[k+1]=0.75×P[k], and the probability value when bin is "0" is theoretically designed as [0.5, 1). If the sliding window parameter a is smaller than 2, the alpha value will appear.

Even if p[k]=1, p[k+1] will be equal to 0.5, not to mention that p[k] will not be equal to 1, so the probability that the bin will be "0" when the sliding window parameter a is 1. Less than 0.5, this is unreasonable, so in this example, the minimum value of the sliding window parameter a is 2.

Assuming that the sliding window parameter a takes a maximum value of 9, the same reason can be calculated that the value of α corresponding to this time is about 0.9980, which is almost close to 1. If the sliding window parameter a is further increased to 10, the corresponding a is about 0.9990. The speed of the probability update at this time has been extremely slow.

In summary, the value range of the sliding window parameter is defined as [2, SWP _max ], where SWP _max represents the maximum value of the sliding window parameter, and this value is not absolutely limited and can be artificially specified. Just guarantee

It falls within the range of (0,1).

In some embodiments, the probability update model of the context is adjusted multiple times based on the number of times the probability of the context is used to encode/decode the binary symbols.

Optionally, in the multiple adjustments to the probability update model, the manner of each adjustment may be the same or different.

For example, each adjustment can be done by increasing the sliding window parameters, and the magnitude of the increase can be the same or different.

For another example, some adjustment methods are to increase the sliding window parameters, and other adjustment methods are to reduce the sliding window parameters.

In some embodiments, in S110, when the number of times the binary symbol is encoded/decoded by the probability update model of the context reaches the threshold, the sliding window parameter of the probability update model is increased; in S120, the sliding is based on the increase The probability update model of the window parameter encodes/decodes the binary symbol; the method further comprises: increasing again when the number of binary symbols for encoding/decoding the binary symbol reaches the threshold again using the adjusted probability update model The probability updates the sliding window parameter of the model; the binary symbol is encoded/decoded based on the probability that the sliding window parameter is increased for the second time.

In some embodiments, the sliding window parameter is again increased by the same magnitude as the last increase.

In some embodiments, the sliding window parameter is again increased by a magnitude that is less than the amplitude of the previous increase.

For example, in the case where the probability state value x of the interval lookup table is smaller than the third value, the amplitude of the sliding window parameter is increased again less than the amplitude of the previous increase.

In some embodiments, the probability update model can be adjusted multiple times during an initialization cycle.

In order to better understand the solution provided by the present application, a specific example of algorithm coding using the solution of the present application is given below.

1) Design context.

The design context, which includes a threshold (denoted as Thr) and a counter (denoted as c). Wherein, the number of bits of the counter can be determined according to the threshold.

For example, when the threshold is set to 64, 128, 256, 510, respectively, the number of bits of the counter should be set to 6, 7, 8, and 9, respectively.

2) Arithmetic coding.

When the encoding bin is "1", the probability update model is:

U[k+1]=U[k]+((2^15-U[k])>>a),V[k+1]=V[k]+((2^15-V[k] )>>b);

When the encoding bin is "0", the probability update model is:

U[k+1]=U[k]-(U[k]>>a), V[k+1]=V[k]-(V[k]>>b);

Final predicted probability: P[k+1]=(U[k+1]+V[k+1])>>1.

Among them, a and b are two sliding window parameters respectively. U and V are two probability update models, respectively. k and k+1 are the current state and the next state, respectively.

It should be understood that at the beginning of step 2), counter c begins counting.

3) Adjustment of sliding window parameters.

The sliding window parameters a and b are adjusted by the solution provided by the present application.

It is assumed that at the time of initialization, the adjustment sliding window parameters a and b are assigned values a0 and b0, respectively.

During the execution of step 2), the counter value of the counter c is monitored in real time. When the counter value of the counter c does not reach the threshold (Thr) designed in step 1), the values of a and b remain unchanged, that is, when the initialization is maintained. Values: a0 and b0; When the count value of the counter c reaches the threshold (Thr) designed in the step 1), the values of a and b are respectively incremented by 1, the value of a becomes a1, and the value of b becomes b1. That is, the adjustment strategies of the sliding window parameters a and b are as follows:

When c<Thr, (a1, b1) = (a0, b0)

When c>Thr, (a1, b1) = (a0+1, b0+1).

It should be understood that there is no absolute sequence relationship between step 2) and step 3) in this example.

It should also be understood that the above examples are illustrative only and not limiting. The adjustment strategies for sliding window parameters a and b can also be as follows:

When c<Thr, (a1, b1) = (a0, b0)

When c ≥ Thr, (a1, b1) = (a0 + Δ1, b0 + Δ2).

When the sliding window parameters a and b are adjusted, the adjustment values Δ1 and Δ2 of the sliding window parameters a and b, respectively, may be the same or different, which is not limited in the present application.

For another example, for the probability update model as shown in formula (3), the solution provided by the present application may also be applied to perform probability update. The specific process is as follows.

1) Design context.

2) Arithmetic coding.

3) Adjustment of sliding window parameters.

The α in the formula (3) can be regarded as a sliding window parameter, and the sliding window parameter α is adjusted according to the count value of the counter c.

It is assumed that the adjustment sliding window parameter α is assigned to α0 at the time of initialization.

During the execution of step 2), the counter value of the counter c is monitored in real time, and according to the technical value of the counter c, the following operations are performed:

If c<=Thr, α'=α;

If c>Thr, α'=α+Δ; where α+Δ should satisfy the value range of (0,1).

Where α denotes the value before the adjustment and α' denotes the value after the adjustment.

It should be noted that the various embodiments described above may be combined in any manner according to intrinsic logic.

In summary, in the embodiment of the present application, when the number of times of coding using the probability update model of the context reaches a threshold, the probability update model is adjusted, and it can be considered that there is already a lot of already encoded information that can be referred to at this time. It is necessary to adjust the probability update model to adapt to the changing coding environment, which can improve the encoding/decoding performance to some extent.

The scheme involving the coding scenario in the method embodiment provided above may be applied to the encoder, and the scheme related to the decoding scenario in the method embodiment provided above may be applied to the decoder.

It should be understood that the solution provided by the embodiments of the present application can be applied to various video codec standards.

The method embodiments of the present application are described above, and the device embodiments of the present application are described below. It should be understood that the description of the device embodiments corresponds to the description of the method embodiments, and therefore, the content that is not described in detail can be referred to the above method implementation. For the sake of brevity, we will not repeat them here.

FIG. 3 is a schematic block diagram of a probability update apparatus 300 for a context provided by an embodiment of the present application. The device 300 includes the following units.

The processing unit 310 is configured to adjust the probability update model when the number of binary symbols that encode/decode the binary symbols using the probability update model of the context satisfies the condition.

The encoding/decoding unit 320 is configured to encode/decode binary symbols based on the adjusted probability update model.

In the solution provided by the present application, the probability update model is adjusted according to the number of times the model is updated according to the probability of using the context, so that the encoding/decoding performance can be improved to some extent.

Optionally, in some embodiments, the number of binary symbols that encode/decode the binary symbols using the probability update model of the context satisfies the condition, including: encoding/decoding the binary symbols when using the probability update model of the context The number of binary symbols reaches the threshold.

In the solution provided by the present application, when the number of times of coding using the probability update model of the context reaches the threshold, the probability update model is adjusted, and it can be considered that there is already a lot of already encoded information that can be referred to at this time, and the probability needs to be adjusted. The model is updated to accommodate the changing coding environment, which can improve the encoding/decoding performance to some extent.

Optionally, in some embodiments, the threshold is a positive integer that is less than a maximum of a probability state value of the arithmetically encoded zoom interval two-dimensional lookup table.

Optionally, in some embodiments, the threshold is adaptively adjusted.

Optionally, in some embodiments, the threshold is determined based on an average number of uses of the probability update model of the same context over a plurality of initialization cycles in which encoding/decoding has been completed.

Optionally, in some embodiments, when the average number of uses is lower than the first value, the threshold is adaptively adjusted to the first value.

Optionally, in some embodiments, when the average number of uses is higher than the second value, the threshold is adaptively adjusted to the second value.

Optionally, in some embodiments, the threshold is an integer between X/2 ⁿ¹ and X/2 ⁿ² , where X is the maximum value of the probability state value of the arithmetically encoded zoom interval two-dimensional look-up table, X/ 2 ^{n1 is} greater than 1, n1 and n2 are natural numbers, and n1 is greater than n2.

Optionally, in some embodiments, the processing unit 310 is specifically configured to: use a counter to count the number of binary symbols that encode/decode the binary symbols using the probability update model of the context; when the counter reaches the counter value At the threshold, it is determined that the number of binary symbols that encode/decode the binary symbols using the probability update model of the context reaches the threshold.

Optionally, in some embodiments, the processing unit 310 is further configured to clear the counter when the current initialization period ends.

Optionally, in some embodiments, the processing unit 310 is further configured to clear the counter when the current stripe slice encoding/decoding ends.

Optionally, in some embodiments, the processing unit 310 is further configured to clear the counter when the current coding tree unit CTU encoding/decoding ends.

Optionally, in some embodiments, the processing unit 310 is specifically configured to adjust the sliding window parameter of the probability update model.

Optionally, in some embodiments, the processing unit 310 is specifically configured to increase a sliding window parameter of the probability update model.

Optionally, in some embodiments, the probability update model includes a plurality of sliding window parameters;

The processing unit 310 is specifically configured to increase the plurality of sliding window parameters, wherein each sliding window parameter increases in amplitude by the same, or different, or not identical.

Optionally, in some embodiments, the probability state value of the arithmetically encoded zoom interval two-dimensional lookup table is less than a third value, and the sliding window parameter increases by less than the fourth value.

Optionally, in some embodiments, the sliding window parameter of the probability update model has a value ranging from 2 to 9.

Optionally, in some embodiments, the processing unit 310 is specifically configured to increase a sliding window parameter of the probability update model. After the encoding/decoding unit 320 encodes/decodes the binary symbols based on the adjusted probability update model, the processing unit 310 is further configured to: when the adjusted probability update model is used to encode/decode binary symbols, When the number of symbols satisfies the second condition, the sliding window parameter of the probability update model is again increased, and the amplitude of the sliding window parameter is increased again or the same as the amplitude of the previous increase. The encoding/decoding unit 320 is further configured to: encode/decode binary symbols based on the probability update model that the sliding window parameter is secondarily increased.

Optionally, in some embodiments, the processing unit 310 is specifically configured to: when the number of binary symbols for encoding/decoding the binary symbols reaches a threshold in the initialization period of the context, when the probability update model of the context is used, The probability update model is adjusted. The encoding/decoding unit 320 is specifically configured to: encode/decode binary symbols based on the adjusted probability update model during the initialization period.

Optionally, in some embodiments, the apparatus further includes: an initializing unit, configured to initialize the sliding window parameter of the probability update model based on a quantization step size of the block to be currently encoded or decoded.

The apparatus 300 can be used to perform the methods provided by the various method embodiments described above.

The functions of the processing unit 310, the encoding/decoding unit 320, and the initialization unit can all be implemented by a processor.

The embodiment of the present application further provides an encoding apparatus for performing the encoding related method in the above method embodiment. The encoder may comprise means for performing the encoding related operations of the above method embodiments.

The embodiment of the present application further provides a decoding apparatus, which is used to perform the decoding related method in the above method embodiment. The encoder may comprise means for performing the decoding related operations of the above method embodiments.

As shown in FIG. 4, an embodiment of the present application further provides an encoder 400, including: a memory 420 for storing instructions, and a processor 410 for executing instructions stored in the memory 420, and Execution of the instructions stored in the memory 420 is such that the processor 410 is operative to perform the encoding related methods of the above method embodiments.

For example, the encoder 400 is a CABAC encoder.

As shown in FIG. 5, the embodiment of the present application further provides a decoder 500, including: a memory 520 for storing instructions, and a processor 510 for executing instructions stored by the memory 520, and Execution of the instructions stored in the memory 520 is such that the processor 510 is operative to perform the decoding related methods of the above method embodiments.

For example, the decoder 500 is a CABAC decoder.

The embodiment of the present application further provides a computer storage medium on which a computer program is stored, and when the computer program is executed by a computer, the computer executes the method of the foregoing method embodiment.

The embodiment of the present application further provides a computer program product comprising instructions, which when executed by a computer, cause a computer to execute the method of the above method embodiment.

In the above embodiments, it may be implemented in whole or in part by software, hardware, firmware or any other combination. When implemented in software, it may be implemented in whole or in part in the form of a computer program product. The computer program product includes one or more computer instructions. When the computer program instructions are loaded and executed on a computer, the processes or functions described in accordance with embodiments of the present invention are generated in whole or in part. The computer can be a general purpose computer, a special purpose computer, a computer network, or other programmable device. The computer instructions can be stored in a computer readable storage medium or transferred from one computer readable storage medium to another computer readable storage medium, for example, the computer instructions can be from a website site, computer, server or data center Transmission to another website site, computer, server or data center via wired (eg coaxial cable, fiber optic, digital subscriber line (DSL)) or wireless (eg infrared, wireless, microwave, etc.). The computer readable storage medium can be any available media that can be accessed by a computer or a data storage device such as a server, data center, or the like that includes one or more available media. The usable medium may be a magnetic medium (for example, a floppy disk, a hard disk, a magnetic tape), an optical medium (such as a digital video disc (DVD)), or a semiconductor medium (such as a solid state disk (SSD)). .

Those of ordinary skill in the art will appreciate that the elements and algorithm steps of the various examples described in connection with the embodiments disclosed herein can be implemented in electronic hardware or a combination of computer software and electronic hardware. Whether these functions are performed in hardware or software depends on the specific application and design constraints of the solution. A person skilled in the art can use different methods to implement the described functions for each particular application, but such implementation should not be considered to be beyond the scope of the present application.

In the several embodiments provided by the present application, it should be understood that the disclosed systems, devices, and methods may be implemented in other manners. For example, the device embodiments described above are merely illustrative. For example, the division of the unit is only a logical function division. In actual implementation, there may be another division manner, for example, multiple units or components may be combined or Can be integrated into another system, or some features can be ignored or not executed. In addition, the mutual coupling or direct coupling or communication connection shown or discussed may be an indirect coupling or communication connection through some interface, device or unit, and may be in an electrical, mechanical or other form.

The units described as separate components may or may not be physically separated, and the components displayed as units may or may not be physical units, that is, may be located in one place, or may be distributed to multiple network units. Some or all of the units may be selected according to actual needs to achieve the purpose of the solution of the embodiment.

In addition, each functional unit in each embodiment of the present application may be integrated into one processing unit, or each unit may exist physically separately, or two or more units may be integrated into one unit.

The foregoing is only a specific embodiment of the present application, but the scope of protection of the present application is not limited thereto, and any person skilled in the art can easily think of changes or substitutions within the technical scope disclosed in the present application. It should be covered by the scope of protection of this application. Therefore, the scope of protection of the present application should be determined by the scope of the claims.

Claims

A context probability update method, comprising:

Adjusting the probability update model when the number of binary symbols that encode/decode binary symbols is satisfied using the probability update model of the context;

The binary symbols are encoded/decoded based on the adjusted probability update model.
The method according to claim 1, wherein the number of binary symbols for encoding/decoding binary symbols using the probability update model of the context satisfies conditions, including:

The number of binary symbols that encode/decode binary symbols reaches a threshold when using the probability update model of the context
The method of claim 2 wherein said threshold is a positive integer that is less than a maximum of a probability state value of a scaled two-dimensional lookup table of an arithmetically encoded scale.
The method of claim 2 wherein said threshold is adaptively adjusted.
The method according to claim 4, wherein the threshold is determined based on an average number of uses of the probability update model of the same context over a plurality of initialization cycles in which encoding/decoding has been completed.
The method of claim 5 wherein said threshold is adaptively adjusted to said first value when said average number of uses is lower than said first value.
The method of claim 5 wherein said threshold is adaptively adjusted to said second value when said average number of uses is higher than said second value.
The method according to any one of claims 2 to 7, wherein the threshold is an integer between X/2 n1 and X/2 n2 , wherein X is an arithmetically coded zoom interval two-dimensional lookup table The maximum value of the probability state value, X/2 n1 is greater than 1, n1 and n2 are natural numbers, and n1 is greater than n2.
The method according to any one of claims 2 to 8, wherein the method further comprises:

Counting the number of binary symbols that encode/decode binary symbols using a probability update model of the context using a counter;

When the count value of the counter reaches the threshold, it is determined that the number of binary symbols that encode/decode binary symbols using the probability update model of the context reaches the threshold.
The method of claim 9 wherein the method further comprises:

The counter is cleared when the current initialization period ends.
The method of claim 9 wherein the method further comprises:

The counter is cleared when the current stripe slice/decode ends.
The method of claim 9 wherein the method further comprises:

The counter is cleared when the current coding tree unit CTU encoding/decoding ends.
The method according to any one of claims 1 to 12, wherein the adjusting the probability update model comprises:

The sliding window parameters of the probability update model are adjusted.
The method according to claim 13, wherein the adjusting the sliding window parameters of the probability update model comprises:

Increase the sliding window parameters of the probability update model.
The method of claim 14, wherein the probability update model comprises a plurality of sliding window parameters;

The increasing the sliding window parameter of the probability update model includes:

The plurality of sliding window parameters are increased, wherein each sliding window parameter increases in amplitude by the same, or different, or not identical.
The method according to claim 14 or 15, wherein the probability state value of the arithmetically encoded zoom interval two-dimensional lookup table is smaller than the third value, and the sliding window parameter increases by less than the fourth value.
The method according to any one of claims 1 to 16, wherein the sliding window parameter of the probability update model has a value ranging from 2 to 9.
The method according to any one of claims 1 to 17, wherein the adjusting the probability update model comprises: increasing a sliding window parameter of the probability update model;

After the encoding/decoding of the binary symbols based on the adjusted probability update model, the method further includes:

When the number of binary symbols for encoding/decoding the binary symbols using the adjusted probability update model satisfies the second condition, the sliding window parameter of the probability update model is again increased, and the sliding window parameter is increased again. The amplitude is the same as or different from the previous increase;

The binary symbol is encoded/decoded based on the probability update model that the sliding window parameter is secondarily increased.
The method according to any one of claims 1 to 17, wherein the probability of updating/decoding a binary symbol by using a probability update model of the context reaches a threshold Update the model to make adjustments, including:

Adjusting the probability update model when the number of binary symbols for encoding/decoding binary symbols reaches a threshold using the probability update model of the context during an initialization period of the context;

The encoding/decoding of the binary symbols based on the adjusted probability update model includes:

The binary symbols are encoded/decoded based on the adjusted probability update model during the initialization period.
The method according to any one of claims 1 to 19, further comprising:

The sliding window parameters of the probability update model are initialized based on the quantization step size of the block to be currently encoded or decoded.
A context probability updating device, comprising:

a processing unit, configured to adjust the probability update model when a quantity of binary symbols for encoding/decoding binary symbols is satisfied by using a probability update model of the context;

And an encoding/decoding unit, configured to encode/decode binary symbols based on the adjusted probability update model.
The apparatus according to claim 21, wherein said number of binary symbols for encoding/decoding binary symbols using the probability update model of said context satisfies conditions, including:

The number of binary symbols that encode/decode binary symbols reaches a threshold when using the probability update model of the context
The apparatus of claim 22 wherein said threshold is a positive integer that is less than a maximum of a probability state value of a scaled two-dimensional lookup table of an arithmetically encoded scale.
The apparatus of claim 22 wherein said threshold is adaptively adjusted.
The apparatus according to claim 24, wherein said threshold is determined based on an average number of uses of a probability update model of the same context over a plurality of initialization periods in which encoding/decoding has been completed.
The apparatus according to claim 25, wherein said threshold is adaptively adjusted to said first value when said average number of uses is lower than said first value.
The apparatus according to claim 25, wherein said threshold is adaptively adjusted to said second value when said average number of uses is higher than said second value.
The apparatus according to any one of claims 22 to 27, wherein the threshold is an integer between X/2 n1 and X/2 n2 , wherein X is an arithmetically encoded zoom interval two-dimensional lookup table The maximum value of the probability state value, X/2 n1 is greater than 1, n1 and n2 are natural numbers, and n1 is greater than n2.
The device according to any one of claims 22 to 28, wherein the processing unit is specifically configured to:

Counting the number of binary symbols that encode/decode binary symbols using a probability update model of the context using a counter;

When the count value of the counter reaches the threshold, it is determined that the number of binary symbols that encode/decode binary symbols using the probability update model of the context reaches the threshold.
The device according to claim 29, wherein said processing unit is further configured to:

The counter is cleared when the current initialization period ends.
The apparatus according to claim 29, wherein said processing unit is further configured to clear said counter when the current stripe slice encoding/decoding ends.
The apparatus according to claim 29, wherein the processing unit is further configured to clear the counter when the current coding tree unit CTU encoding/decoding ends.
The apparatus according to any one of claims 21 to 32, wherein the processing unit is specifically configured to adjust a sliding window parameter of the probability update model.
The apparatus according to claim 33, wherein the processing unit is specifically configured to increase a sliding window parameter of the probability update model.
The apparatus according to claim 34, wherein said probability update model comprises a plurality of sliding window parameters;

The processing unit is specifically configured to increase the plurality of sliding window parameters, wherein each sliding window parameter increases in amplitude by the same, or different, or not identical.
The apparatus according to claim 34 or 35, wherein the probability state value of the arithmetically encoded zoom interval two-dimensional look-up table is smaller than the third value, and the sliding window parameter increases by less than the fourth value.
The apparatus according to any one of claims 21 to 36, wherein the sliding window parameter of the probability update model has a value ranging from 2 to 9.
The apparatus according to any one of claims 21 to 37, wherein the processing unit is specifically configured to increase a sliding window parameter of the probability update model;

After the encoding/decoding unit encodes/decodes the binary symbols based on the adjusted probability update model, the processing unit is further configured to: encode/decode the binary symbols when the adjusted probability update model is used. When the number of binary symbols satisfies the second condition, the sliding window parameter of the probability update model is again increased, and the amplitude of the sliding window parameter is increased again or the same as the amplitude of the previous increase;

The encoding/decoding unit is further configured to encode/decode binary symbols based on the probability update model that the sliding window parameter is secondarily increased.
The apparatus according to any one of claims 21 to 37, wherein the processing unit is configured to: when the initialization period of the context is used, edit a binary symbol when using a probability update model of the context The probability update model is adjusted when the number of decoded binary symbols reaches a threshold;

The encoding/decoding unit is specifically configured to: encode/decode binary symbols based on the adjusted probability update model during the initialization period.
The device according to any one of claims 21 to 39, wherein the device further comprises:

And an initializing unit, configured to initialize a sliding window parameter of the probability update model based on a quantization step size of a block to be currently encoded or decoded.
An encoder/decoder, comprising: a memory for storing instructions, said processor for executing instructions stored by said memory, and performing execution of instructions stored in said memory The processor is used to perform the method of any one of claims 1 to 20.
A computer storage medium, characterized in that a computer program is stored thereon, the computer program being executed by a computer such that the computer performs the method of any one of claims 1 to 20.
A computer program product comprising instructions, wherein the instructions, when executed by a computer, cause a computer to perform the method of any one of claims 1 to 20.