JP2010171655A - Arithmetic coding apparatus, and arithmetic coding method - Google Patents

Abstract

<P>PROBLEM TO BE SOLVED: To make coding efficiency satisfactory, and to easily initialize a context memory. <P>SOLUTION: During initializing the context memory 114 for storing information between an advantageous symbol (valMPS) and occurrence probability (pStateIdx) of the advantageous symbol of each context in each slice, a combination of valMPS and pStateIdx of each context calculated from three tables is compared with a combination of valMPS and pStateIdx of each context upon the end of coding of the preceding slice stored in the context memory 114 to generate evaluation values. Between the generated evaluation values, a combination of valMPS and pStateIdx to have the best evaluation value is determined as an initial value. <P>COPYRIGHT: (C)2010,JPO&INPIT

Description

  The present invention relates to an arithmetic encoding device, an arithmetic encoding method, a program, and a storage medium, and particularly to a technique suitable for use in initializing a context memory.

  In recent years, with advances in digital signal processing, high integration and high speed of LSI, a large amount of digital information such as moving images, still images, and voices can be efficiently encoded and recorded on a recording medium, or by a communication medium It has become possible to perform transmission. Recently, digital high-definition broadcasting and terrestrial digital broadcasting have been started by applying such technology. In addition, the performance of digital video cameras and the like has been improved, and photographed images have been improved in image quality.

  As an image coding standard, H.264, which has been formulated by ITU-T in recent years. H.264 and the like are attracting attention (see, for example, Non-Patent Document 1). This H. In H.264, encoding is realized with higher efficiency than MPEG which is a conventional video encoding standard. H. One of the H.264 implementations of highly efficient encoding is entropy encoding. As a method for realizing high-efficiency entropy coding, H.264 is used. In H.264, CABAC (Context-based Adaptive Binary Arithmetic Coding) using arithmetic coding can be used.

  However, in the CABAC system, it is necessary to initialize a storage device called a context memory that has addresses from 0 to 459 at the head of a slice and holds 7-bit data for one address. Note that since the context memory is initialized at the time when encoding of one slice is completed, encoding reflecting the property of the image being encoded cannot be performed at the head of the slice.

  As a method of initializing the context memory, for example, the context memory is initialized using three initialization variable tables. For example, in the arithmetic encoding device described in Patent Document 1, encoding efficiency is increased by appropriately selecting and using an initialization variable table. Specifically, first, initialization is performed using the initialization variable in the first initialization variable table, and encoding is performed until a predetermined number of bits is generated. Next, initialization is performed using the initialization variable in the second initialization variable table, and encoding is similarly performed until a predetermined number of bits is generated. Finally, initialization is performed using the initialization variable of the third initialization variable table, and similarly, encoding is performed until a predetermined number of bits is generated, and the most efficient initialization variable table is selected, Encoding for one slice is performed.

JP 2006-279333 A ITU-T Recommendation H. H.264

  However, the arithmetic coding apparatus described in Patent Document 1 has a problem that it takes a lot of processing time because it encodes each of the three initialization variable tables by a predetermined number of bits. .

  An object of the present invention is to improve the coding efficiency and to easily initialize a context memory in view of the above-mentioned problems.

  The arithmetic coding apparatus according to the present invention initializes, for each slice, a context memory that stores information on a dominant symbol for each context and the occurrence probability of the dominant symbol, and the dominant symbol and occurrence probability stored in the context memory. An arithmetic encoding device that performs arithmetic encoding of binary data using the information of the storage unit, the storage unit storing information of a plurality of combinations of dominant symbols and occurrence probabilities for each context, and stored in the storage unit An evaluation means for generating an evaluation value by comparing a plurality of combinations of the dominant symbols and occurrence probabilities with a combination of the dominant symbols and occurrence probabilities for each context at the end of encoding of the previous slice, and generated by the evaluation means Based on the evaluated value, the dominant symbol for storing in the context memory at the time of initialization And characterized in that it has an initial value determining means for determining an initial value of the probability.

  The arithmetic coding method of the present invention initializes, for each slice, a context memory that stores information on the dominant symbol for each context and the occurrence probability of the dominant symbol, and the dominant symbol and occurrence probability stored in the context memory. Is an arithmetic coding method for performing binary data arithmetic coding using a plurality of combinations of dominant symbols and occurrence probabilities stored in the storage means, and each context at the end of the previous slice encoding. An evaluation step for generating an evaluation value by comparing a combination of the prevailing symbol and the occurrence probability, and a preferential symbol to be stored in the context memory at the time of initialization based on the evaluation value generated in the evaluation step And an initial value determining step for determining an initial value of the occurrence probability.

  The program of the present invention initializes, for each slice, a context memory that stores information on the dominant symbol for each context and the probability of occurrence of the dominant symbol, and also stores information on the dominant symbol and occurrence probability stored in the context memory. A program that causes a computer to execute each step of an arithmetic encoding method that performs arithmetic encoding of binary data using a plurality of combinations of dominant symbols and occurrence probabilities stored in storage means, and a code of a previous slice An evaluation step for generating an evaluation value by comparing a combination of the dominant symbol and occurrence probability for each context at the end of the initialization, and in the context memory at the time of initialization based on the evaluation value generated in the evaluation step A dominant symbol for storing and an initial value determining step for determining an initial value of an occurrence probability. Characterized in that to execute the Yuta.

  The storage medium of the present invention stores the program described above.

  According to the present invention, when the context memory is initialized, the encoding efficiency is improved, and the context memory can be easily initialized without the need for encoding.

(First embodiment)
Hereinafter, embodiments of the present invention will be described with reference to the drawings.
First, the CABAC encoding process will be described with reference to FIG. FIG. 2 is a block diagram showing a configuration example of the CABAC arithmetic coding apparatus 200 according to the present embodiment.
In FIG. 2, the CABAC method arithmetic coding apparatus 200 includes a binarization unit 201, a context calculation unit 202, a context memory 203, a binary arithmetic coding unit 204, and a context memory initialization unit 205.

  The binarization unit 201 inputs information D21 such as a differential motion vector, orthogonal transform coefficient, prediction mode, etc., performs binarization processing defined by the standard, and generates a binarized symbol (binary data) D22 To do. The details of the binarization process are described in Non-Patent Document 1.

  The context memory 203 stores 460 pieces of information on the most probable symbol (hereinafter, valMPS) having a high occurrence probability and the occurrence probability (hereinafter, pStateIdx) of the dominant symbol for each context index. The context calculation unit 202 calculates a context index D23 based on the binarized symbol D22. Then, the valMPS and pStateIdx information D24 is read from the context memory 203 using the context index D23 as an address, and output to the binary arithmetic encoding unit 204.

  The context memory initialization unit 205 receives the context index D23 and initializes the context memory 203. The initialization of the context memory 203 will be described later in the description of FIG.

  The binary arithmetic coding unit 204 performs arithmetic coding processing based on the binarized symbol D22 and the information D24 of valMPS and pStateIdx, and outputs the result D26 to the outside. The binary arithmetic encoding unit 204 performs arithmetic encoding for each binarized symbol, and outputs and updates valMPS and pStateIdx information D25 to the context memory 203 each time. Thereby, the encoding suitable for the property of the image being encoded is realized.

Next, the operation of the binary arithmetic encoding unit 204 will be described in more detail with reference to FIG.
In FIG. 3, a binary arithmetic encoding unit 204 performs arithmetic encoding processing using variables called codIRange and codILow that specify regions. In arithmetic coding, a number line is divided by the number of values that a symbol can take. Since binary arithmetic coding is used here, the number line is divided into two. The size of each area at the time of division is determined based on the probability of occurrence of each symbol.

  According to the example shown in FIG. 2, first, the valMPS and pStateIdx information D24 read from the context memory 203 based on the context index D23 is input. At this time, the width of codIRange is determined based on the input pStateIdx information. The codIRange of the region at time t1 shown in FIG. 3 indicates the width of the symbol region having a high probability of occurrence, and codILow indicates the lower limit value of the region. Note that pStateIdx takes a value from 0 to 63, and the closer the value is to 63, the greater the width of codIRange.

  At time t1, the binarized symbol D22 output from the binarization unit 201 is input bit by bit. At this time, if the input binarized symbol matches valMPS, the entire region at time t2 is set to the width of codIRange.

  Next, at time t2, new pStateIdx and valMPS information D24 are input, and the region is divided in the same manner. In the example shown in FIG. 3, the region 301 is codIRange at time t2, and the lower limit 302 is codILow at time t2. Similarly, the next binarized symbol is input bit by bit, and if the binarized symbol and valMPS are different, the region 303 is selected as the entire region at time t3. Similarly at time t3, the region is divided based on pStateIdx. The lower limit 304 is codILow at time t3, and the region 305 is codIRange at time t3.

  If the division is repeated in this manner and binary arithmetic coding proceeds, the area becomes narrower and the area cannot be expressed. Therefore, in the CABAC method, when the width of codIRange becomes smaller than 0x100, normalization processing (hereinafter referred to as Renorm processing) is performed to maintain expression accuracy. Also, when performing the Renorm process, a code is output based on the value of codILow. The details of this operation are described in Non-Patent Document 1.

  In binary arithmetic coding, if valMPS and the binarized symbol do not match and codIRange is rapidly reduced, the frequency of occurrence of the Renorm processing is increased accordingly, and more codes are generated. Therefore, if the frequency with which the valMPS matches the binarized symbol is high, the occurrence frequency of the Renorm process is reduced, the generation of codes is suppressed, and the encoding efficiency is improved. Therefore, in the CABAC system, valMPS and pStateIdx are updated every time one bit of a binarized symbol is encoded, so that the valMPS and the binarized symbol easily coincide with each other as the encoding proceeds. As a result, encoding efficiency is improved.

Next, the context memory initialization operation will be described with reference to FIGS.
FIG. 4 is a diagram illustrating a relationship among variables (m, n), a context index (hereinafter referred to as “ctxIdx”), and cabac_init_idc used for initialization of the context memory described in Non-Patent Document 1. In FIG. 4, ctxIdx indicates 11 to 23, but ctxIdx is defined from 0 to 459.

  cabac_init_idc is included in the slice header and is used when the context memory is initialized. For example, if the value of cabac_init_idc in the slice header is 0 at the beginning of the slice, initialization is performed using m and n in the row where cabac_init_idc is 0 in FIG.

A conventional context memory initialization operation will now be described with reference to FIG. FIG. 5 is a block diagram showing a configuration example for initializing a conventional context memory.
In FIG. 5, when initializing the context memory, the first initialization variable table 511, the second initialization variable table 512, the third initialization variable table 513, the selection unit 501, An initial value generation unit 502 and a context memory 503 are used.

  In the first initialization variable table, information on initialization variables m and n when cabac_init_idc is 0 is held for ctxIdx. The second initialization variable table holds information about initialization variables m and n when cabac_init_idc is 1, for ctxIdx. Similarly, in the third initialization variable table, information on initialization variables m and n when cabac_init_idc is 2 is stored for ctxIdx.

  The ctxIdx generation unit 504 generates a signal from 0 to 459 and outputs the signal to the first initialization variable table 511, the second initialization variable table 512, and the third initialization variable table 513. When ctxIdx is input to the first initialization variable table 511, the second initialization variable table 512, and the third initialization variable table 513, the selection unit 501 uses the ctxIdx as an address as an initialization variable m. And n information is read from each table. Here, D50 is information on the initialization variables m and n output from the first initialization variable table 511, and D51 is information on the initialization variables m and n output from the second initialization variable table 512. It is. D52 is information on initialization variables m and n output from the third initialization variable table 513.

  When the selection unit 501 inputs the information D50 to D52 of the initialization variables m and n, the initialization variable information D53 corresponding to a predetermined value is output to the initial value generation unit 502 according to the value of cabac_init_idc included in the slice header. Note that ctxIdx may be supplied only to the initialization variable table corresponding to the value of cabac_init_idc in the slice header.

  The initial value generation unit 502 inputs the values of m and n from the input information D53 of the initialization variables m and n into the formula shown in the following formula 1, and further uses the formula shown in the following formula 2 to The values of valMPS and pStateIdx for storing in are calculated.

  Here, Clip3 (x, y, z) is Clip3 (x, y, z) = x when z <x, and Clip3 (x, y, z) = y when z> y. In other cases, Clip3 (x, y, z) = z. Also, a different quantization parameter (SliceQPy) (not shown) is input for each slice, and an initial value is generated.

  As described above, the valMPS and pStateIdx values for ctxIdx460 are calculated, and the valMPS and pStateIdx information D54 is output to the context memory 503 having addresses 0 to 459 and holding 7 bits of data for one address. Further, ctxIdx (context index) D55 is output as the write address of the context memory 503. The context memory is initialized by the above processing.

  The context memory initialization operation described above is always performed at the head of the slice. Here, various methods can be considered as to which initialization variable table is used to initialize the context memory from the first initialization variable table to the third initialization variable table. For example, when an initialization variable table is selected using the method described in Patent Document 1, encoding is performed for a predetermined number of bits, which requires a lot of processing time. Therefore, in the present embodiment, a procedure for determining a suitable initialization variable table more easily will be described.

FIG. 1 is a block diagram illustrating a configuration example of the context memory initialization unit 205 according to the present embodiment.
In FIG. 1, the context memory initialization unit 205 includes a first initialization variable table 101, a second initialization variable table 102, and a third initialization variable table 103. In each of the initialization variable tables 101 to 103, ctxIdx and information on initialization variables m and n for each cabac_init_idc are stored.

  The first initial value generation unit 104 receives a different quantization parameter (SliceQPy) for each slice, and information D11 of initialization variables m and n for each context index (ctxIdx) from the first initialization variable table 101. To do. In addition, the second initial value generation unit 105 inputs the quantization parameter (SliceQPy) and the information D12 of the initialization variables m and n for each context index (ctxIdx) from the second initialization variable table 102. Similarly, the third initial value generation unit 106 inputs a quantization parameter (SliceQPy) and information D13 of initialization variables m and n for each context index (ctxIdx) from the third initialization variable table 103. .

  Then, the first initial value generation unit 104 calculates valMPS and pStateIdx for each ctxIdx using the equations shown in Equation 1 and Equation 2, and stores the information D14 of valMPS and pStateIdx in the first initial value storage unit 107. Output and store. The second initial value generation unit 105 and the third initial value generation unit 106 calculate in the same manner, and information D15 and D16 of valMPS and pStateIdx, respectively, are stored in the second initial value storage unit 108 and the third initial value storage, respectively. The data is output to the unit 109 and stored.

  The first evaluation unit 110 receives the valMPS and pStateIdx information D25 stored in the context memory 114 at the time when the previous slice has been encoded. Further, the information D17 of valMPS and pStateIdx is read from the first initial value storage unit 107. Next, an absolute difference value between valMPS and pStateIdx at the time when encoding of the previous slice is completed and valMPS and pStateIdx stored in the first initial value storage unit 107 is obtained for each ctxIdx. Then, the evaluation value based on the absolute difference value is determined, and the evaluation value information D20 is output to the initial value determination unit 113. Similarly, the second evaluation unit 111 and the third evaluation unit 112 read the information D18 and D19 of valMPS and pStateIdx, respectively, and output the information D21 and D22 of the evaluation values determined to the initial value determination unit 113, respectively. To do.

  The initial value determination unit 113 selects the initial value (valMPS and pStateIdx) of the evaluation value having the highest evaluation from the three input evaluation values. Then, the selected initial value information D23 and the corresponding ctxIdxD24 are output to the context memory 114.

  The first initialization variable table 101 to the third initialization variable table 103 are realized by a storage device such as a ROM. Also, as shown in FIG. 4, when any value of 0, 1, 2 in cabac_init_idc and the value of ctxIdx from 0 to 459 are given, initialization variable m corresponding to cabac_init_idc and ctxIdx And n information is read out.

  The first evaluation unit 110 to the third evaluation unit 112, at the end of the previous slice or the start of the current slice, first, information on valMPS and pStateIdx having a value of ctxIdx held in the context memory 114 of 0 Enter. Then, valMPS and pStateIdx information each having a ctxIdx value of 0 stored in each of the first initial value storage unit 107 to the third initial value storage unit 109 are input. Note that valMPS and pStateIdx held in the context memory 114 are valMPS and pStateIdx at the time when encoding of the previous slice is completed.

  Next, the first evaluation unit 110 calculates the absolute difference between the two input pStateIdx. At the same time, the second evaluation unit 111 and the third evaluation unit 112 also calculate the same absolute difference value. Then, points are added in descending order of the three difference absolute values. Here, the added point is an evaluation value of a penalty method as a penalty, and a large evaluation value indicates that a difference between the initial value and the final state of the previous slice is large. However, when valMPS is different, a certain predetermined point is added. Such evaluation is performed with the value of ctxIdx from 0 to 459, and the evaluation value information D20 to D22 of each cabac_init_idc is output to the initial value determination unit 113.

  The initial value determination unit 113 receives the evaluation value information D20 to D22, and finally generates the initial values of valMPS and pStateIdx having the smallest (good) evaluation values. The initial values of the valMPS and pStateIdx of the initialization variable table Select. Then, the context memory 114 is initialized using the initial value. Specifically, the initial value information D23 of valMPS and pStateIdx using cabac_init_idc having the best evaluation value is output together with ctxIdxD24 and stored in the context memory 114. As a result, the values closest to valMPS and pStateIdx updated in the previous slice can be inherited, and encoding suitable for the properties of the image can be performed from the head of the slice. In addition, since encoding is not performed using the initial value of each cabac_init_idc, the time for selecting the optimum cabac_init_idc is also reduced.

FIG. 6 is a flowchart illustrating an example of an operation procedure for each context of the context memory initialization unit 205 according to the present embodiment.
First, in step S61, the first evaluation unit 110 determines whether or not it is the head of a slice. When determining whether or not it is the head of a slice, if there is a change in the value of SliceQPy that differs for each slice, it may be determined that it is the head of a slice. If the result of this determination is that it is the head of a slice, the process proceeds to step S62. On the other hand, if the result of determination in step S61 is not the head of the slice, the process waits as it is. The same process is performed in the second evaluation unit 111 and the third evaluation unit 112.

  Next, in step S62, the first evaluation unit 110 uses the initial value information of the dominant symbol (valMPS) calculated from the first initialization variable table 101 and the occurrence probability (pStateIdx) as the first initial value. Read from the storage unit 107. Then, the information of the dominant symbol (valMPS) and the occurrence probability (pStateIdx) at the end of encoding of the previous slice stored in the context memory 114 is input. Then, the two pStateIdx are compared, and the absolute value of the difference is calculated. The same process is performed in the second evaluation unit 111 and the third evaluation unit 112.

  Next, in step S63, the first evaluation unit 110 to the third evaluation unit 112 determine an evaluation value based on the difference absolute value calculated in step S62. Specifically, three difference absolute values calculated in each of the first evaluation unit 110 to the third evaluation unit 112 are compared, and a penalty is added as an evaluation value to the largest value. A penalty is also added to the next largest difference evaluation value. Then, the evaluation value calculated in this way is temporarily held, and the process proceeds to the next step S64.

  Next, in step S64, it is determined whether or not the value of ctxIdx has been counted up to 459. This is a necessary step for calculating all evaluation values in ctxIdx. As a result of this determination, if ctxIdx is less than 459, the process proceeds to step S65, and if ctxIdx is 459, the process proceeds to step S66.

  Next, in step S65, the value of ctxIdx is incremented, and the process proceeds to step S62. On the other hand, in step S66, the initial value determination unit 113 selects the lowest (best) evaluation value among the evaluation values for each cabac_init_idc obtained by the processing from step S61 to step S65. Then, the dominant symbol (valMPS) and its occurrence probability (pStateIdx) are initialized using the selected evaluation value cabac_init_idc, and the initialization process is terminated.

  Note that the timing at which the first initial value generation unit 104 calculates the initial value of the dominant symbol and occurrence probability for each cabac_init_idc may be performed in step S62, and is stored in advance before becoming the head of the slice. May be.

  As described above, according to this embodiment, since the evaluation value is calculated from each table and the best combination is selected, the initial values of the dominant symbol and the occurrence probability close to the final state of the previous slice are obtained. It is done. As a result, the values closest to valMPS and pStateIdx updated in the previous slice can be inherited, and encoding suitable for the properties of the image can be performed from the head of the slice.

  In this embodiment, the initial value of the dominant symbol and the occurrence probability close to the final state of the previous slice is used at the time of initialization. However, when there is a scene change, the image characteristics are It may be very different from the final state of the slice. Therefore, when a scene change is detected, the processing in this embodiment may be stopped and the initial value may be determined by a conventional method.

  In this embodiment, the initial value of the dominant symbol and the occurrence probability close to the final state of the previous slice is used at the time of initialization. However, the previous slice of the same type than the previous slice of a different type. It may be better to use. In this case, if the slice to be encoded is, for example, a P slice, the dominant symbol and occurrence probability of the P slice of the same type closest in time may be used.

(Other embodiments according to the present invention)
Each means constituting the arithmetic coding apparatus and each step of the arithmetic coding method in the embodiment of the present invention described above can be realized by operating a program stored in a RAM or ROM of a computer. This program and a computer-readable storage medium storing the program are included in the present invention.

  In addition, the present invention can be implemented as, for example, a system, apparatus, method, program, storage medium, or the like. Specifically, the present invention may be applied to a system including a plurality of devices. The present invention may be applied to an apparatus composed of a single device.

  The present invention includes a case where a software program (in the embodiment, a program corresponding to the flowchart shown in FIG. 6) for realizing the functions of the above-described embodiments is directly or remotely supplied to a system or apparatus. This includes the case where the system or the computer of the apparatus is also achieved by reading and executing the supplied program code.

  Accordingly, since the functions of the present invention are implemented by computer, the program code installed in the computer also implements the present invention. In other words, the present invention includes a computer program itself for realizing the functional processing of the present invention.

  In that case, as long as it has the function of a program, it may be in the form of object code, a program executed by an interpreter, script data supplied to the OS, and the like.

  Examples of the storage medium for supplying the program include a flexible disk, a hard disk, an optical disk, and a magneto-optical disk. Further, there are MO, CD-ROM, CD-R, CD-RW, magnetic tape, nonvolatile memory card, ROM, DVD (DVD-ROM, DVD-R) and the like.

  As another program supply method, there is a method of connecting to a homepage on the Internet using a browser of a client computer. It can also be supplied by downloading the computer program itself of the present invention or a compressed file including an automatic installation function from a homepage to a storage medium such as a hard disk.

  It can also be realized by dividing the program code constituting the program of the present invention into a plurality of files and downloading each file from a different homepage. That is, the present invention includes a WWW server that allows a plurality of users to download a program file for realizing the functional processing of the present invention on a computer.

  As another method, the program of the present invention is encrypted, stored in a storage medium such as a CD-ROM, distributed to users, and encrypted from a homepage via the Internet to users who have cleared predetermined conditions. Download the key information to be solved. It is also possible to execute the encrypted program by using the key information and install the program on a computer.

  Further, the functions of the above-described embodiments are realized by the computer executing the read program. Furthermore, based on the instructions of the program, an OS or the like running on the computer performs part or all of the actual processing, and the functions of the above-described embodiments can be realized by the processing.

  As another method, a program read from a storage medium is first written in a memory provided in a function expansion board inserted into the computer or a function expansion unit connected to the computer. Then, based on the instructions of the program, the CPU or the like provided in the function expansion board or function expansion unit performs part or all of the actual processing, and the functions of the above-described embodiments are also realized by the processing.

It is a block diagram which shows the structural example of the context memory initialization part which concerns on embodiment of this invention. It is a block diagram which shows the structural example of the arithmetic encoding apparatus of the CABAC system in embodiment of this invention. It is a conceptual diagram which shows an example of the operation | movement of binary arithmetic coding. It is a figure which shows the relationship between the variable (m, n) used for initialization of a context memory, a context index, and cabac_init_idc. It is a block diagram which shows the structural example for initializing the conventional context memory. It is a flowchart which shows an example of the operation | movement procedure of the context memory initialization part which concerns on embodiment of this invention.

101 First initialization variable table 102 Second initialization variable table 103 Third initialization variable table 104 First initial value generation unit 105 Second initial value generation unit 106 Third initial value generation unit 107 1 initial value storage unit 108 second initial value storage unit 109 third initial value storage unit 110 first evaluation unit 111 second evaluation unit 112 third evaluation unit 113 initial value determination unit 114 context memory 201 2 Valuation unit 202 context calculation unit 203 context memory 204 binary arithmetic encoding unit 205 context memory initialization unit

Claims (9)

A context memory that stores information on the dominant symbol for each context and the probability of occurrence of the dominant symbol is initialized for each slice, and binary data is stored using the information on the dominant symbol and the probability of occurrence stored in the context memory. An arithmetic encoding device that performs arithmetic encoding,
Storage means for storing information of a plurality of combinations of dominant symbols and occurrence probabilities for each context;
Evaluation means for generating an evaluation value by comparing a plurality of combinations of dominant symbols and occurrence probabilities stored in the storage means with a combination of dominant symbols and occurrence probabilities for each context at the end of encoding of the previous slice; ,
And an initial value determining means for determining an initial value of a dominant symbol and an occurrence probability to be stored in the context memory at the time of initialization based on the evaluation value generated by the evaluation means. Encoding device.   The initial value determining means determines, as an initial value to be stored in the context memory, a combination having the best evaluation value among a plurality of combinations of dominant symbols and occurrence probabilities stored in the storage means. The arithmetic coding apparatus according to claim 1.   The evaluation means calculates a difference absolute value between the occurrence probability of the dominant symbol stored in the storage means and the occurrence probability of the dominant symbol at the end of encoding of the previous slice, and calculates the difference absolute value The arithmetic coding device according to claim 1, wherein an evaluation value is generated based on the arithmetic coding device.   The arithmetic coding apparatus according to claim 1, wherein the evaluation unit adds a weight to the evaluation value according to a result of comparison between the dominant symbol and the occurrence probability.   If the type of the previously encoded slice and the type of the next encoded slice are different, the evaluation means calculates the evaluation value using the dominant symbol and occurrence probability of the slice that is the closest in time. The arithmetic coding apparatus according to claim 1, wherein the arithmetic coding apparatus is generated.   2. The evaluation unit according to claim 1, wherein the evaluation unit does not generate an evaluation value when a scene change exists between a previously encoded slice and a next encoded slice. Arithmetic coding device. A context memory that stores information on the dominant symbol for each context and the probability of occurrence of the dominant symbol is initialized for each slice, and binary data is stored using the information on the dominant symbol and the probability of occurrence stored in the context memory. An arithmetic encoding method for performing arithmetic encoding,
An evaluation step for generating an evaluation value by comparing a plurality of combinations of dominant symbols and occurrence probabilities stored in the storage means with a combination of dominant symbols and occurrence probabilities for each context at the end of encoding of the previous slice;
And an initial value determining step of determining an initial value of a dominant symbol and an occurrence probability to be stored in the context memory at the time of initialization based on the evaluation value generated in the evaluation step. Encoding method. A context memory that stores information on the dominant symbol for each context and the probability of occurrence of the dominant symbol is initialized for each slice, and binary data is stored using the information on the dominant symbol and the probability of occurrence stored in the context memory. A program for causing a computer to execute each step of an arithmetic encoding method for performing arithmetic encoding,
An evaluation step for generating an evaluation value by comparing a plurality of combinations of dominant symbols and occurrence probabilities stored in the storage means with a combination of dominant symbols and occurrence probabilities for each context at the end of encoding of the previous slice;
An initial value determining step for determining a dominant symbol and an initial value of occurrence probability to be stored in the context memory at the time of initialization based on the evaluation value generated in the evaluation step is performed. Program.   A computer-readable storage medium storing the program according to claim 8.

Images