JP2007142637A - Image information encoder - Google Patents

Abstract

<P>PROBLEM TO BE SOLVED: To provide an image information encoder in which the amount of codes generated by CABAC (Context-Adaptive Binary Arithmetic Coding) can be prevented from exceeding a maximum amount of codes with less computational complexity without requiring an extra variable length encoding module. <P>SOLUTION: The variable length encoding section 100 comprises a section for quantizing data about moving picture and then arithmetically encoding binarized data, a section 106 for calculating the amount of generated codes by counting the number of bits having a determined value and the digit places where the value is not determined in the arithmetically encoded data from the arithmetically encoding section 105, and predicting whether a predetermined upper limit of the amount of generated codes is exceeded or not when the image of 1MB is arithmetically encoded from the amount of generated codes calculated for image data of predetermined unit, and a section 103 for rewriting the binarized data to a value having an amount of bits smaller than an original value when it is predicted that the upper limit is exceeded. <P>COPYRIGHT: (C)2007,JPO&INPIT

Description

  The present invention relates to an image information encoding device that encodes image information of a moving image, and more particularly to an image information encoding device that outputs a quantized DCT coefficient or the like after variable length encoding.

  In the moving picture coding technique, the coding efficiency is improved. As a result, it is becoming possible to realize a smooth-moving videophone or to shoot a high-quality moving image on a mobile phone. With the progress of such coding technology, H.264 / MPEG-4 AVC, which is the international standard for the latest video compression coding technology, specifies that DCT coefficients and motion vectors are transmitted in syntax. As an entropy coding method for syntax elements, a coding method with higher coding efficiency is prepared instead of conversion by a simple table. That is, a variable-length coding method called CAVLC (Context-Adaptive Variable Length Coding) and an arithmetic coding method called CABAC (Context-Adaptive Binary Arithmetic Coding). Of these, arithmetic coding methods such as CABAC are said to be able to theoretically compress stationary signals to the limit. FIG. 9 is a diagram illustrating an example of conventional response processing when the generated code amount by CABAC exceeds the standard maximum code amount. CABAC has the advantage of good coding efficiency as described above, but has the disadvantage that the amount of generated code is not known until the processing is completed. In H.264 / AVC, there is an upper limit of 3200 bits for the amount of generated code per macroblock (hereinafter referred to as MB). However, in CABAC, the amount of generated code is unknown until the processing is completed. Whether or not the standard maximum code amount is satisfied cannot be determined in advance. For this reason, when the generated code amount exceeds the maximum code amount, the CABAC process must be performed again while changing the encoding condition such as quantization as shown in FIG. 9A.

On the other hand, in CAVLC, since the amount of generated code can be predicted according to the amount of input binary data, CABAC and CAVLC are provided in parallel, and the generated amount of CABAC is maximum. In the case where the code amount is not satisfied, a conventional technique is disclosed in which a code string is output after switching to variable length coding by CAVLC (see Patent Document 1).
JP 2004-135251 A

  However, starting CABAC from the quantization stage involves a complicated process of restoring the state transition of the context calculation unit that updates the table while calculating the occurrence probability of the code in time series. There is a problem of requiring a large amount of calculation.

  In addition, providing CABAC and CAVLC in parallel always provides a circuit that is normally unnecessary in preparation for exceeding the amount of generated code that rarely occurs. As a result, there is a problem in that the cost and size of the image information encoding device are increased.

  In view of the above problems, an object of the present invention is to provide an image information code that can reduce the amount of code generated by CABAC from exceeding the maximum amount of code with a smaller amount of calculation without providing an extra variable length coding module. It is to provide a device.

  In order to solve the above problems, an image information encoding device of the present invention is an image information encoding device that arithmetically encodes binarized data and outputs a code string, and after quantizing data relating to a moving image, By calculating the arithmetic encoding means for arithmetically encoding the binarized data, and the number of bits in which the value is determined and the digit position where the value is not determined in the arithmetically encoded data by the arithmetic encoding means When a certain image is arithmetically encoded from the generated code amount calculation means for calculating the generated code amount and the generated code amount calculated for a predetermined unit of image data, a predetermined generation for the image is generated. Code amount prediction means for predicting whether or not the upper limit value of the code amount is exceeded, and writing that rewrites the binarized data to a value having a smaller bit amount than the original value when predicted to exceed the upper limit value With replacement means .

  The present invention can be realized not only as such an image information encoding device, but also as a context adaptive binary arithmetic encoding device provided in such an image information encoding device. Such image information encoding apparatus can be realized as an image information encoding method having steps as characteristic means, or can be realized as a program for causing a computer to execute these steps. Needless to say, such a program can be distributed via a recording medium such as a CD-ROM or a transmission medium such as the Internet.

  As described above, according to the present invention, the generated code amount can be suppressed within the maximum code amount determined by the standard without performing the CABAC process again.

(Embodiment)
FIG. 1 is a block diagram showing a configuration of variable length coding section 100 of the present embodiment. The variable-length encoding unit 100 predicts the generated code amount for 1 MB during CABAC processing for each MB, and binarizes when it is about to exceed the maximum code amount of 3200 bits defined for 1 MB in the H.264 standard. A variable-length encoding unit that rewrites a data value to a value with a smaller bit amount, including a binarization processing unit 101, a binary buffer 102, a binary data rewriting unit 103, a context calculation unit 104, an arithmetic An encoding unit 105, a code amount prediction unit 106, a binary data rewrite control unit 107, and a buffer 108 are provided. The binarization processing unit 101 converts, for example, quantized data represented by multiple values such as a decimal number into a binary number, and outputs the generated binary data to the context calculation unit 104 and the binary buffer 102. The binary buffer 102 accumulates data binarized by the binarization processing unit 101. The binarized data rewriting unit 103 reads the binarized data from the binarized buffer 102 in accordance with an instruction from the binarized data rewriting control unit 107 described later, and rewrites it to a value with a smaller number of bits. The rewritten data is output to the context calculation unit 104 and the arithmetic coding unit 105. The context calculation unit 104 recalculates the occurrence probability of the code of the binary data input from the binarized data rewriting unit 103, and updates the probability table. The arithmetic encoding unit 105 arithmetically encodes the binary data input from the binary data rewriting unit 103 with reference to the probability table updated by the context calculation unit 104, and outputs the determined code to the buffer 108 To do. The code amount prediction unit 106 monitors the encoding result from the arithmetic encoding unit 105 and determines whether or not the MB code amount exceeds the maximum code amount. The binarized data rewrite control unit 107 holds a list of bit amounts (not shown) to be reduced from the binarized data according to the prediction result by the code amount prediction unit 106, and the code amount prediction unit 106 If it is determined from the prediction result that the maximum code amount of 1 MB is exceeded, the binarized data rewriting unit 103 is controlled to rewrite the binarized data to a value with a smaller bit amount.

  Hereinafter, prediction of the generated code amount will be described in more detail. Hereinafter, a method for monitoring the code amount during arithmetic coding and predicting the code amount of the MB being processed from the bit amount will be described. At this time, since the code value may not be determined, the code amount can be known by referring to a counter value called Bits_Outstanding. FIG. 2 is a block diagram showing a CABAC (Encode Decision) 200 that is a detailed configuration of the context calculation unit 104 and the arithmetic coding unit 105 shown in FIG. A CABAC (Encode Decision) 200 that performs CABAC includes an interval / probability table update unit 207, an interval renormalization unit 201, and a code value output processing unit 206. The section / probability table update unit 207 stores a context variable table in which a context index ctxidx, a ratio of occurrence probability, and a symbol (0 or 1) having a higher occurrence probability are associated with each other, and is input from the outside. Based on the context index ctxidx, the occurrence probability of 1 and 0 of the digit is read from the context variable table. Furthermore, in a section where the lower limit coordinate is 0 or more and the upper limit coordinate value is 1 or less, codeIRange representing the upper limit coordinate value of the section already obtained for the input binary data, The codeILow representing the lower limit coordinate value is output as Range information to the interval renormalization unit 201. The interval renormalization unit 201 counts a range determination module 202, a module 203 that outputs a code value = 0, a module 204 that outputs a code value = 1, and which digit code value is not fixed. A counter Bits_Outstanding205 is provided. The range determination module 202 is input with the digit to be encoded of binary data that is an array of 0 and 1 via the section / probability table update unit 207. Further, codeIRange representing the upper limit coordinate value of the section and codeILow representing the lower limit coordinate value of the section are input. The range determination module 202 uses the occurrence probability of the occurrence of (0, 1) of each digit given from the context calculation unit 104 (that is, the interval / probability table update unit 207) as the width of the interval. Multiplication is performed, and the multiplication result is added to the lower limit coordinate value or subtracted from the upper limit coordinate value, thereby reducing the width of the section. Thereby, a binary value that can be taken within the width of the section is determined. In other words, if the actually generated bit is 0, the Range determination module 202 multiplies the interval width (codeIRange-codeILow) by the occurrence probability of 0 in that digit. A value obtained by adding this multiplication result to the lower limit codeILow of the section is set as a new upper limit codeIRange of the section. In this case, the lower limit coordinate value codeILow is left as it is. Conversely, if the actually generated bit is 1, the Range determination module 202 multiplies the occurrence probability of 1 in that digit by the width of the section (codeIRange-codeILow), and subtracts the multiplication result from the codeIRange at that time To do. A value obtained as a result of the subtraction is set as a new lower limit coordinate value codeILow of the section. In this case, the upper limit coordinate value codeIRange is left as it is. In this way, the Range determination module 202 calculates a real value range from 0 to 1 given to the already generated sequence of 0 and 1 for each digit of the input binary data. When the range of this real value is less than a certain range, it is determined whether the binary value that can be taken after the decimal point is 0 or 1. For example, when the lower limit coordinate value codeILow of the section is 0.5 or more, it is determined that the first decimal place value is 1. Range determination module 202 selects module 203 that outputs code value = 0 if the determined code value is 0, and selects module 204 that outputs code value = 1 if the determined code value is 1. To do. When the code value is not fixed, Bits_Outstanding205 is selected. The confirmed code value is output as a code value via the code value output processing unit. The module 203 for outputting the code value = 0, the module 204 for outputting the code value = 1, and the Bits_Outstanding 205, after outputting each signal, the Range determination module to update the Range in order to calculate the code value of the next bit. A signal is output to 202. Therefore, the code amount prediction unit 106 can know the code amount after CABAC without waiting for the determination of the code value (0, 1) by monitoring at the data amount monitoring position shown in FIG.

  In the present embodiment, in monitoring the code amount in CABAC, the CABAC code amount is monitored by monitoring Bits_Outstanding 205 which is an indeterminate stage, instead of counting those determined as code values. In CABAC, the code value is determined by the probability table, but the code value is not determined by the state of the establishment table and is counted in Bits_Outstanding205. Therefore, in order to know the code amount, there is no need to wait for the code value to be determined. By monitoring this, the code amount can be accurately monitored.

  The binary data rewrite control unit 107 changes binary data that is input data for arithmetic coding when the maximum code amount of 1 MB of code is likely to be exceeded, and changes the code amount after arithmetic coding thereafter. Suppress. The code amount after arithmetic coding is unknown at the time of input, but if the bit amount of input data is small, the bit amount of output data is also small.

  FIG. 3 is a diagram illustrating a configuration example in the case of changing the value by going back to the quantization process (not shown) in the previous stage of the variable length coding unit 100 in FIG. As shown in the figure, the quantization / binarization processing timing control module 301 is configured so that the determination result of the range determination module 202 in the interval renormalization unit 201 of the CABAC processing shown in FIG. The generated code amount is calculated, and when the 1 MB code amount is likely to exceed the maximum code amount 3200 bits, the binarized data binarized by the binarization processing module 303 is changed to a smaller value. At this time, the amount of generated code after arithmetic coding is suppressed by retroactively changing to the quantized data that is the input of the binarization processing. At this time, the code amount can be suppressed by forcibly reducing all the quantized coefficients that have not yet been binarized or a part of the quantized coefficients. Here, quantization or binarization is not performed again, but the quantization coefficient is directly rounded down or rounded. Also, the quantization coefficient may be rewritten to a predetermined value.

  It is also possible to measure the amount of code at the time when a certain syntax is input, and to change subsequent quantized data or binary data when the measured amount of code is equal to or greater than a certain value. Fig. 4 shows binary data by comparing the ratio B of the output data of binary data in 1MB with the ratio A of the generated code quantity in the maximum code quantity for 1MB when a certain syntax is input. It is a figure which shows an example which determines the change of digitization data. For example, as shown on the left side of the figure, it is assumed that luminance DC and luminance AC are output from the binary data buffer with a certain syntax syntax as the head. The ratio B of this output data is 50% of the bit amount of the binarized data for 1 MB, and the ratio A of the code amount generated by the arithmetic coding is 75% with respect to the maximum code amount for 1 MB. Suppose. At this time, the binarized data rewrite control unit 107 calculates (A−B), and changes the binarized data after the color difference DC when the value of (A−B) exceeds a predetermined threshold. It may be determined to do so. Further, as described above, when the binarized data rewrite control unit 107 determines that the generated code amount for 1 MB exceeds the maximum code amount when a certain number of syntax elements are input, the following 2 The binarization process may be omitted by replacing with a default value without performing the binarization. At that time, as shown in FIG. 3, by controlling the timing of quantization and binarization, rewriting can be performed without unnecessary quantization and binarization.

  The binary data syntax elements include MB type, motion vector, quantization step, pixel information, and the like. The pixel information described here indicates difference information between the predicted image and the original image in the encoding process. This pixel information includes luminance DC, luminance AC, color difference DC, and color difference AC. Pixel information and motion vectors are dominant in the amount of code after arithmetic coding. Therefore, the code amount after arithmetic coding can be suppressed by changing the pixel information and the motion vector. Note that the binarized data rewriting unit 103 and the quantization / binarization processing timing control module 301 may not change the binary data that is the target of processing that does not perform compression called “Bypass” of arithmetic coding. The coefficient suffix is exponential Golomb encoded and not CABAC. This is called bypass. Of course, it goes without saying that binary data called bypass of arithmetic coding may be changed according to circumstances. Depending on the case, for example, a large amount of code needs to be reduced, and a certain coefficient is set to 0 for both prefix and suffix.

  In addition, changing the pixel information as described above causes image quality degradation. Therefore, in order to minimize the influence, priority is provided to the pixel information change target. 5A and 5B show the order of the syntax elements of the image information input from the binary data buffer shown in FIG. 4 to the arithmetic coding module, and the change of the binary data for each syntax element. It is a figure which shows the priority. As shown in FIG. 5A, the syntax elements of each MB are (1) binary data of luminance DC, (2) binary data of luminance AC, (3) binary data of color difference DC, and (4) Binary data of color difference AC is input to the arithmetic coding module in order. On the other hand, as shown in FIG. 5 (b), the change target of the binarized data of the pixel information includes (1) an AC component of color difference, (2) an AC component of luminance, (3) a DC component of color difference, (4) Prioritize and change the order of luminance DC components. That is, DC binary data is protected as much as possible. For pixel information with a low priority, for example, (1) AC component of the color difference component or the like, the value is greatly changed such as truncation, and for pixel information with a high priority, for example, (4) a luminance DC component, etc. Do not change the value or change the value by reducing the change width. By changing the priority order and the value change range for each component in combination, image quality deterioration can be made less noticeable. Here, it is also possible to select and change the frequency component to be changed from the feature amount of the image coding information.

  As an example of the priority, a high priority component is a low frequency component of pixel information, and a low priority component is a high frequency component of pixel information.

  Similarly, the value of the quantization coefficient can also be changed in this manner by setting priorities for each syntax element.

  In addition to the above priorities, the binarized data rewrite control unit 107 assigns priorities to binary data changes according to the positions of blocks in the MB.

  FIG. 6 is a diagram showing the position of a 4 × 4 pixel block to be protected as much as possible from rewriting binary data in the MB in the case of in-plane prediction, with respect to luminance AC and color difference AC. In the in-plane prediction, MB is made up of 16 × 16 pixels and divided into 16 small blocks of 4 × 4 pixels or 4 small blocks of 8 × 8 pixels for prediction processing.

  The MB to be changed may be referenced from other adjacent MBs. Since other MBs refer to the pixel information of the MB currently being changed when decoding H.264, if the binary data of the pixel information in the reference portion is changed, the other MBs are decoded. Sometimes image quality degradation can propagate. That is, since the pixel information of the rightmost row and the lowest column of the MB to be changed is referred to, the pixel information of the block including the pixel is not changed as much as possible even for the AC component.

  Specifically, in inter prediction and intra 4 × 4 prediction, as shown in FIG. 6, the pixel information at the positions of block numbers 5, 7, 10, 11, 13, 14, and 15 of luminance AC is not changed as much as possible. Further, the pixel information at the positions of the block numbers 1, 2, and 3 of the color difference AC is not changed as much as possible. Similarly, in the intra 8 × 8 and the intra 16 × 16, the pixel information of the small block including one row at the right end of the MB and one column at the bottom is not changed.

  However, information on where to be referred to is acquired and held in advance before DCT, and if block numbers 5, 7, 13, and 15 are not referenced, they are added to the coefficient change candidates. Similarly, when block numbers 10, 11, 14, and 15 are not referred to, they are added to the coefficient change candidates. The coefficient change candidate is similarly changed in the 4 × 4 block of the color difference AC.

  In the example shown in FIG. 4, the binarized data rewrite control unit 107 generates a ratio B of binary data output from the binary data buffer and a code generated by arithmetically encoding the output binary data. The ratio A of the amount to the 1 MB maximum code amount is compared, and it is determined that the binarized data is changed when the difference between A and B exceeds a predetermined threshold. However, the present invention is not limited to this, and the change of the binarized data may be determined according to another method. For example, a model curve indicating the time change of the generated code amount by CABAC and the change width of the corresponding binarized data is calculated in advance, and from the time change of the generated code amount for a certain amount of binarized data. The necessity of changing the binarized data and the degree of change may be determined.

  FIG. 7 is a diagram illustrating a method of predicting an increase in the subsequent code amount from a code amount that has been arithmetically encoded using a modeled increase curve in the MB that is the target of arithmetic encoding. The model curves (1) to (4) shown in FIG. 7 are assumed to be an increase curve modeled from an increase curve of the code amount measured from actual data. If it is determined by this prediction that the standard maximum code amount is exceeded, the binary data is changed by the above method.

  That is, when the arithmetic encoding of the constant syntax element is completed, the generated code amount at that time is compared with a model curve as shown in FIG. More specifically, the prediction is made from the arithmetic and coding results for the first 1 to 3 sub-blocks of syntax and luminance DC or syntax, luminance DC and luminance AC. Thereby, the data change amount of the binarized data is determined according to the model curve that most closely approximates the increase rate of the generated code amount. For example, in the case of the model curve (1), the generated code amount suddenly increases immediately after the start of arithmetic coding of the luminance DC, and reaches the code amount dangerous area. Accordingly, in this MB, it can be estimated that the excess code amount exceeding the standard maximum code amount of 1 MB is large. Therefore, a large change amount of pixel information is set. As an example, it is possible to change the coefficient of pixel information after reaching the code amount risk area to 0 or the like. Conversely, in the case of model (3) and the like, it can be estimated that the excess code amount exceeding the standard maximum code amount of 1 MB is small, so the change amount of the pixel information is set small.

  Normally, the code amount control is performed by changing a quantization parameter or the like, but in the present invention, the binary data after execution of quantization is changed to control the code amount. Since the quantization parameter is known at the time of arithmetic coding, the fact that deterioration in image quality is not noticeable even if the binary data is changed is used at the time of rough quantization, and the binary data according to the quantization parameter. It is also possible to determine the amount of change.

  When changing binary data, the value of coeff_abs_level_minus1 is changed for pixel information, and the value of mvd is changed for a motion vector. At this time, since the value to be changed is a coefficient after DCT, the high frequency component is preferentially reduced according to the code amount control amount. When changing the value of the binarized data, in order to rewrite the data at a desired position, an address is assigned to the binary buffer 102 and stored so that the address corresponds to the position of the pixel information. It may be. FIG. 8A is a diagram illustrating a state of the binary buffer 102 that stores the position and address of the pixel information of the 4 × 4 pixel block in the MB to be changed of the binary data in association with each other. Here, taking the case of inter-plane prediction of 4 × 4 pixel blocks as an example, one word is allocated and stored for each block index in a memory area of 18 bits and 1 word. FIG. 8B is a diagram showing an example of binarized data for rewriting of a 4 × 4 pixel block stored in a rewriting memory (not shown). As shown in FIG. 8A, syntax is shown in the leftmost column of the binary buffer 102, and binary data is stored in the order of block_index, coded_block_flag, last_significant_flag, coeff_abs_level_minus1_0 to coeff_abs_level_minus1_15. ing. block_index indicates which block in the MB is the binarized data, and the uppermost block_index is 0, which indicates that the block is block number 0 shown in FIG. In this way, by assigning block_index to data, it is possible to manage which block's coefficient position is the binarized data stored below. coded_block_flag is a coded block flag, and last_significant_flag is a final significant coefficient flag. The following coeff_abs_level_minus1 is one element of a variable related to a macroblock of the H.264 standard, and binary data can be changed in the same manner as mvd. coeff_abs_level_minus1 means a value obtained by subtracting 1 from the absolute value of the prediction error after quantization. coeff_abs_level_minus1 is different from mvd in that it is only a prefix when the value is 0-14, and when coeff_abs_level_minus1 is a value greater than 15, the prefix is 0-14 and the suffix is 15 or greater. For example, when coeff_abs_level_minus1 = 20, prefix = 14, suffix = 20−14 = 6, and prefix = 14 is CABAC. Again, both prefix and suffix are changed. Coeff_abs_level_minus1_0 to coeff_abs_level_minus1_15 store binarized data that is coefficient information of each pixel of the block. As shown in FIG. 8A, by storing the binarized data in the binary buffer 102, when coefficient rewriting occurs, by reading from the rewriting address instead of the corresponding address, the desired data position can be obtained. Rewriting can be performed efficiently.

  In FIG. 8B, the binarized data shown in FIG. 8A is determined in advance to change the coefficient for each block of 4 × 4 pixels in order to reduce the amount of generated code by arithmetic coding. The coefficient sequence of the binarized data obtained is shown. As shown in the figure, the coefficient 0 is not stored for coeff_abs_level_minus1_0 to coeff_abs_level_minus1_15 in the replacement memory. If the coefficient is 0, it is necessary to change the last_significant_coeff_flag, so ceff_abs_level_minus1 is changed to 1 or more so that the last_significant_coeff_flag does not change. By doing in this way, there is an effect that the overhead concerning change processing can be reduced.

  Further, mvd indicates one element of a variable related to the MB of the H.264 standard. This represents the difference between pmv, which is the median of motion vectors (MV) of adjacent MBs, and the MV of the encoding target MB. That is, mvd (x, y) = MV (x, y) −pmv (x, y).

As pre-processing for arithmetic coding of mvd, first, mvd is binarized. The binarization method is different between the prefix part which is the object of arithmetic coding and the suffix part which is not the object of arithmetic coding. That is, the suffix part is not compressed. For this reason, in this embodiment, it is possible to reduce the MV code amount by deleting the suffix part. For example, when the value of MV is up to 9, the prefix is set, and after that the suffix is set. At this time,
(Example 1) When mvd = 8, encoding is performed with a prefix of 8.
(Example 2) When mvd = 17, encoding is performed with a prefix of 9 and a suffix of 17-9 = 8.

  As described above, in Example 1, mvd is compressed, but in Example 2, the suffix part 8 is not compressed. Therefore, by deleting this suffix part, the amount of code after arithmetic coding can be reduced. In the case of Example 2, the suffix part 8 may be changed to 7. Even in the case of Example 1, the prefix may be changed to 0 when it is necessary to greatly reduce the code amount.

  In the above-described embodiment, (1) the color difference AC component, (2) the luminance AC component, (3) the color difference DC component, and (4) the luminance change for the binarized data representing the pixel information. Although priority is given in the order of DC components, the present invention is not limited to this method. First, the binarized data representing the color difference is changed, and then the binarized data representing the luminance is changed. A ranking may be set. For example, (1) color difference AC component, (2) color difference DC component, (3) luminance AC component, and (4) luminance DC component may be arranged in this order.

  Each functional block in the block diagrams (FIGS. 1, 2, 3, etc.) is typically realized as an LSI that is an integrated circuit. These may be individually made into one chip, or may be made into one chip so as to include a part or all of them. (For example, the functional blocks other than the memory may be integrated into one chip.)

  The name used here is LSI, but it may also be called IC, system LSI, super LSI, or ultra LSI depending on the degree of integration.

  Further, the method of circuit integration is not limited to LSI's, and implementation using dedicated circuitry or general purpose processors is also possible. An FPGA (Field Programmable Gate Array) that can be programmed after manufacturing the LSI or a reconfigurable processor that can reconfigure the connection and setting of circuit cells inside the LSI may be used.

  Further, if integrated circuit technology comes out to replace LSI's as a result of the advancement of semiconductor technology or a derivative other technology, it is naturally also possible to carry out function block integration using this technology. Biotechnology can be applied.

  In addition, among the functional blocks, only the means for storing the data to be encoded or decoded may be configured separately instead of being integrated into one chip.

  The image information encoding apparatus according to the present invention is useful as an image information encoding apparatus provided in a personal computer, a PDA, a digital broadcast station, a mobile phone, and the like having a communication function.

It is a block diagram which shows the structure of the variable length encoding part of this Embodiment. It is a block diagram which shows the arithmetic coding module which is a detailed structure of the context calculation part enclosed with the broken line in FIG. 1, and the arithmetic coding part. It is a figure which shows the structural example in the case of changing a value retroactively to the quantization process which is not shown in the front | former stage of the variable length encoding part of FIG. When a certain syntax is input, change the binarized data by comparing the ratio B of output data of binary data in 1MB with the ratio A of the generated code amount in the maximum code amount for 1MB. It is a figure which shows an example which determines. (A) and (b) are the order of the syntax elements of the image information inputted from the binary data buffer shown in FIG. 4 to the arithmetic coding module, and the priority of the change of the binary data for each syntax element. It is a figure which shows a ranking. In the case of in-plane prediction, it is a figure which shows the position of the block of 4x4 pixel which should be protected as much as possible from rewriting of binarized data in MB regarding luminance AC and color difference AC. It is a figure which shows the method of predicting the increase of the subsequent code amount from the code amount already arithmetically encoded using the modeled increase curve in the MB to be arithmetically encoded. (A) is a figure which shows the state of the binary buffer which matches and memorize | stores the position and address of the pixel information of 4x4 pixel block in MB which becomes the change object of binary data. (B) is a figure which shows an example of the binarization data for rewriting of the 4x4 pixel block stored in the rewriting memory which is not shown in figure. It is a figure which shows an example of the conventional response processing when the generated code amount by CABAC exceeds the standard maximum code amount.

Explanation of symbols

DESCRIPTION OF SYMBOLS 100 Variable length coding part 101 Binary processing part 102 Binary buffer 103 Binary data rewriting part 104 Context calculation part 105 Arithmetic coding part 106 Code amount prediction part 107 Binary data rewriting control part 108 Buffer 200 CABAC (Encode Decision)
201 Section Renormalization Unit 202 Range Determination Module 203 Module that Outputs Code Value = 0 204 Module that Outputs Code Value = 1 205 Bits_Outstanding
206 Code value output processing unit 207 Interval / probability table update unit 301 Quantization / binarization processing timing control module 302 Quantization module 303 Binary processing module

Claims (14)

An image information encoding apparatus that arithmetically encodes binary data and outputs a code string,
Arithmetic encoding means for arithmetically encoding binarized data after quantizing the data relating to the moving image;
In the arithmetic encoded data by the arithmetic encoding means, the generated code amount calculating means for calculating the generated code amount by counting the number of bits whose value is determined and the digit position where the value is not determined;
A code that predicts whether or not a predetermined image upper limit of the generated code amount is exceeded when a certain image is arithmetically encoded from the generated code amount calculated for the predetermined unit of image data Quantity prediction means;
An image information encoding device comprising: rewriting means for rewriting the binarized data to a value having a smaller bit amount than the original value when predicted to exceed the upper limit value. The image information encoding apparatus according to claim 1, wherein the rewriting means further rewrites the quantized data before binarization in addition to the binarized data. The image information encoding apparatus according to claim 1, wherein the rewriting unit changes binarized data of pixel information representing luminance and color difference of a pixel. The image information encoding according to claim 1, wherein the rewriting means rewrites the quantized data before the binarization of the pixel information in addition to the binarized data of the pixel information representing the luminance and color difference of the pixel. apparatus. The rewriting unit rewrites the binarized data in the order of color difference AC, luminance AC, color difference DC, and luminance DC when rewriting the pixel information representing the luminance and color difference of the pixel. Image information encoding device. The rewriting unit does not rewrite pixel information of a pixel referred to from another block for a pixel in a certain image that has been subjected to in-plane encoding. Image information encoding device. The arithmetic coding means is CABAC,
The generated code amount calculation means calculates the current generated code amount from the number of output bits of the module that outputs the determined code value in CABAC and the BITS OUTSTANDING that counts the number of bits of the digit for which the code value is not fixed,
The code amount prediction means predicts whether or not the total code amount generated for a macroblock that is a target of CABAC exceeds the upper limit value based on the calculated number of output bits,
The pixel image information encoding apparatus according to claim 1, wherein the rewriting means rewrites binary data that is an input of CABAC. The image recoding device according to claim 7, wherein the rewriting means rewrites the binarized data that is an input of CABAC and the quantized data before binarization. The image information encoding device further includes:
An address is assigned to each block of a predetermined size for each syntax element, and a binarized data buffer that stores binarized data input to CABAC for each block;
A replacement memory storing in advance binary data with a small bit amount for each block of the size for each syntax element;
9. The image information encoding according to claim 8, wherein the rewriting means rewrites the binarized data of the block to be rewritten with the binarized data stored in the reread memory for each block. apparatus. The rewriting means rewrites the binarized data of the motion vector in addition to the pixel information, and rewrites the value of coeff_abs_level_minus1 if the rewriting target is pixel information, and if the rewriting target is a motion vector The image information encoding device according to claim 7, wherein the value of mvd is rewritten. The rewriting means rewrites the pixel information indicated by coeff_abs_level_minus1 to binary data having a smaller bit amount than the original value when changing the binary data of the pixel information, and does not change last_significant_coeff_flag. The image information encoding device according to claim 7. An image information encoding apparatus that arithmetically encodes binary data and outputs a code string,
After quantizing the data relating to the moving image, the binary data is arithmetically encoded while counting the number of bits for which the value has been determined and the digit position where the value has not been determined. When the upper limit value of the generated code amount determined in advance is predicted when arithmetic coding is performed, and when it is predicted that the upper limit value is exceeded, the binarized data is converted into the bit amount of the original value. Image information encoding device that rewrites to a smaller value. An image information encoding method for arithmetically encoding binary data and outputting a code string,
An arithmetic encoding step of arithmetically encoding the binarized data after quantizing the data relating to the moving image;
A generated code amount calculating step of calculating a generated code amount by counting the number of bits whose value is determined and the digit whose value is not determined, among the data being arithmetically encoded;
A code that predicts whether or not a predetermined image upper limit of the generated code amount is exceeded when a certain image is arithmetically encoded from the generated code amount calculated for the predetermined unit of image data A quantity prediction step;
A rewriting step of rewriting the binarized data to a value having a smaller bit amount than the original value when it is predicted that the upper limit value will be exceeded. An integrated circuit that arithmetically encodes binary data and outputs a code string,
Arithmetic encoding means for arithmetically encoding binarized data after quantizing the data relating to the moving image;
In the arithmetic encoded data by the arithmetic encoding means, the generated code amount calculating means for calculating the generated code amount by counting the number of bits whose value is determined and the digit position where the value is not determined;
A code that predicts whether or not a predetermined image upper limit of the generated code amount is exceeded when a certain image is arithmetically encoded from the generated code amount calculated for the predetermined unit of image data Quantity prediction means;
An integrated circuit comprising: rewriting means for rewriting the binarized data to a value having a smaller bit amount than the original value when predicted to exceed the upper limit value.

Images