JP6549057B2 - Encoding apparatus and encoding method - Google Patents

Description

  Embodiments of the present invention relate to an encoding apparatus and an encoding method.

  For example, when transmitting various signals such as image, voice, video, sensor data, and other one-dimensional or multi-dimensional digital signals, in general, when the signal information amount is larger than the communication capacity of a predetermined transmission path, By encoding these signals, transmission is performed with a reduced amount of information. In this case, as a coding method, for example, various types of processing such as prediction coding, transform coding, quantization, or variable length coding, or processing using any two or more of these in combination is generally used. Among them, when the variable length coding method is used, the code amount generated when a predetermined processing unit is coded may not be constant. In this case, generally, a configuration is used in which a reception buffer is provided in the previous stage of the decoding device and the encoded signal (code) input from the encoding device through the transmission path is temporarily stored. In such a configuration, a reception buffer having a capacity large enough for the fluctuation range of the code amount of the code output from the encoding device is prepared, and the code amount of the code stored in the reception buffer is equal to that of the reception buffer. After being accumulated up to a predetermined rate, the decoding process is started. This makes it possible to suppress a shortage of codes accumulated in the reception buffer, that is, occurrence of buffer underflow, at the timing when the decoding apparatus starts decoding.

  Also, in general, when the capacity of the reception buffer is increased, it is possible to suppress the buffer underflow even if the fluctuation range of the code amount of the code output from the coding apparatus is increased. On the other hand, since the time (initial delay time) until the reception buffer is first accumulated to a predetermined ratio of the capacity becomes long, there is a problem that the delay time of the entire transmission system becomes long. Therefore, from the viewpoint of the initial delay time, it is desirable that the capacity of the reception buffer be as small as possible.

  Based on the above viewpoints, as a method of suppressing the occurrence of buffer underflow while suppressing the capacity of the reception buffer, the fluctuation of the code amount is suppressed by controlling the code amount output by the encoding device, and the occurrence of buffer underflow How to prevent is known. As an example of such a method, for example, in video coding, a method of encoding a frame group to be encoded twice is known. In this method, the amount of codes to be output is controlled by determining the quantization parameter used in the second encoding using the statistic value collected in the first encoding, and the occurrence of buffer underflow is suppressed. There is. Also, as an example of another method, when the code amount when encoding a certain processing unit exceeds the allocated code amount, the method of suppressing the generated code amount by skipping the encoding of the exceeded part It is also known.

  However, in the above-described method of encoding a frame group twice, there is a problem that the amount of processing required for the encoding process is large because the encoding process needs to be performed twice. In addition, since the quantization parameter is determined based on the estimated value of the code amount and encoding is performed, there is a possibility that a deviation may occur between the estimated value and the actually generated code amount, and the reception buffer If it is small, there is a possibility that the occurrence of buffer underflow can not be suppressed.

  Further, in the method of suppressing the generated code amount by skipping the above-mentioned coding, there is a problem that the quality of the decoded signal is largely deteriorated in the portion where the coding is skipped.

Patent No. 5096342 JP 2001-169281 A

  The present invention has been made in view of the above, and it is an object of the present invention to provide an encoding apparatus and encoding method that can suppress buffer underflow with a small amount of processing.

The encoding apparatus according to the embodiment is an encoding apparatus that performs variable-length encoding on a quantized value obtained by quantizing an input signal based on a quantization parameter, and includes a selection unit, an acquisition unit, and a determination unit. , A quantization unit, and an encoding unit. The selection unit selects one variable-length coding table based on the characteristics of the input signal. An acquisition unit acquires the maximum code amount that can be generated for each of the quantization parameters according to the selected variable length coding table. The determination unit determines the quantization parameter based on the maximum code amount for each quantization parameter and the allocated code amount determined by the occupied code amount. The quantization unit quantizes the input signal using a quantization parameter to obtain a quantization value. The coding unit selects one variable-length coding table based on the quantization parameter, and performs variable-length coding on the quantized value.

FIG. 1 is a diagram showing an example of the entire configuration of a transmission system according to a first embodiment. It is a figure which shows an example of the hardware constitutions of the encoding apparatus which concerns on 1st Embodiment. It is a figure which shows an example of the block configuration of the encoding apparatus based on 1st Embodiment. It is a figure explaining control of the occupancy code amount of a receiving buffer. It is a figure which shows an example of the graph which shows the worst code amount with respect to each quantization parameter. It is a flowchart which shows an example of the encoding process of the encoding apparatus which concerns on 1st Embodiment. It is a figure which shows an example of the block configuration of the encoding apparatus which concerns on 2nd Embodiment.

  Hereinafter, an encoding apparatus and an encoding method according to each embodiment of the present invention will be described in detail with reference to the drawings. However, since the drawings are schematic, specific configurations should be determined in consideration of the following description.

First Embodiment
FIG. 1 is a diagram showing an example of the entire configuration of the transmission system according to the first embodiment. The entire configuration of a transmission system 1 according to the present embodiment will be described with reference to FIG.

  As shown in FIG. 1, the transmission system 1 according to the present embodiment includes an encoding device 10, a decoding device 20, and a reception buffer 30. A signal (code stream to be described later) output from the encoding device 10 is accumulated in the reception buffer 30 via the network 40. The transmission path transmitted from the encoding device 10 to the reception buffer 30 is not limited to a wired transmission path such as the network 40, and a signal is transmitted from the encoding device 10 to the network 40 by wireless communication. It is good also as a thing. Further, although the configuration of the transmission system 1 shown in FIG. 1 shows a configuration in which the decoding device 20 and the reception buffer 30 are separate, the present invention is not limited to this. For example, the decoding device 20 May include the reception buffer 30, and the reception buffer 30 may be disposed at the previous stage of the decoding process by the decoding device 20.

  The encoding apparatus 10 is an apparatus that encodes an input signal input from the outside and outputs the code to the outside (for example, the network 40). Specifically, the encoding device 10 quantizes the input signal, and performs variable-length coding on a quantized value, which is a quantized signal, and outputs the result.

  Here, examples of the input signal include voice, image, video, three-dimensional volume data, output signals of various sensors (one-dimensional or multi-dimensional signals), and transmission signals (baseband signals) in a wireless or wired transmission path. Etc. Note that the input signal is not limited to any of the above-mentioned signals. For example, various signals having a small entropy with respect to the number of input bits, or a signal whose entropy is reduced by predictive coding or transform coding described later May be used as the input signal.

  For example, as preprocessing for an input signal, the encoding device 10 predicts the input signal based on close samples, obtains a difference value (prediction residual) from the predicted value (prediction processing), and inputs this prediction residual The signal may be quantized and variable-length coded (so-called predictive coding). For example, in the case of an output signal of a temperature sensor or the like, signals which are temporally close to each other are likely to have similar values, and it is possible to reduce the amount of information (entropy) by prediction processing.

  Also, the encoding device 10 performs transformation processing such as discrete Fourier transformation, discrete cosine transformation, or discrete wavelet transformation on the input signal or the above-mentioned prediction residual, and quantizes and changes the converted signal as the input signal. It may be long coded (so-called transform coding). For example, in an image signal, particularly in a natural image, the signal tends to be concentrated in a low spatial frequency region, and by performing frequency conversion processing, it is possible to reduce the entropy of the high spatial frequency region. In the case of image signals, it is generally known how to reduce entropy more efficiently by combining prediction processing and conversion processing.

  In this manner, entropy is reduced by performing appropriate prediction processing or conversion processing according to the input signal, and high-efficiency coding is realized by variable-length coding this signal (or its quantized value). Ru. In the present embodiment, for convenience, an encoding apparatus will be described in which an input signal with a low entropy (for example, prediction residual) with respect to the number of input bits is assumed, and this is quantized and variable-length encoded and output.

  The decoding device 20 is a device that withdraws a code of a predetermined unit from the reception buffer 30 in which the encoded signal output from the encoding device 10 is accumulated, to the decoding timing for each predetermined time, and decodes the withdrawn code. is there.

  The reception buffer 30 is a storage device that accumulates the encoded signal output from the encoding device 10 at a constant rate. Further, the reception buffer 30 is disposed at the previous stage of the decoding function of the decoding device 20, and a predetermined unit of code is taken out for decoding by the decoding device 20 at every predetermined time of decoding timing.

  The network 40 is a wireless or wired transmission path connecting the encoding device 10 and the reception buffer 30.

  FIG. 2 is a diagram illustrating an example of a hardware configuration of the encoding device according to the first embodiment. The hardware configuration of the encoding device 10 according to the present embodiment will be described with reference to FIG.

  As shown in FIG. 2, the encoding device 10 according to the present embodiment includes an input I / F 501, an encoding processor 502, a cache 502a, a random access memory (RAM) 503, an output I / F 504, and an auxiliary And a storage device 505. The input I / F 501, the encoding processor 502, the RAM 503, the output I / F 504, and the auxiliary storage device 505 are communicably connected to each other by a bus 506 such as an address bus and a data bus.

  An input I / F 501 is an interface for inputting an input signal from the outside.

  The encoding processor 502 is an arithmetic unit that controls the overall operation of the encoding device 10 shown in FIG. The encoding processor 502 performs an encoding process.

  The cache 502 a is a memory that temporarily stores data when the encoding processor 502 performs processing as a work area.

  The RAM 503 is a volatile storage device that temporarily stores a code stream and the like generated by the encoding process of the encoding processor 502. The RAM 503 is, for example, an SRAM (Static RAM) or a DRAM (Dynamic RAM).

  The output I / F 504 is an interface for outputting the code stream generated by the encoding process of the encoding processor 502 to the outside.

  The auxiliary storage device 505 is a storage device for storing a variable length coding table and the like described later. The auxiliary storage device 505 is a storage device capable of storing electrically, magnetically or optically, such as a hard disk drive (HDD), a solid state drive (SSD), a flash memory or an optical disk. Although the auxiliary storage device 505 is assumed to be included in the encoding device 10, the present invention is not limited to this, and may be installed outside the encoding device 10, and in this case, the auxiliary storage device The data stored in 505 may be communicated to the encoding device 10 via an interface. Further, the information stored in the above-described auxiliary storage device 505 may be incorporated as a processing method (logic circuit or program) instead of being stored in the auxiliary storage device 505.

  The hardware configuration of the encoding device 10 shown in FIG. 2 is an example, and may be realized by another configuration.

  FIG. 3 is a diagram showing an example of a block configuration of the coding apparatus according to the first embodiment. FIG. 4 is a diagram for explaining control of the occupied code amount of the reception buffer. The block configuration and operation of the encoding device 10 according to the present embodiment will be described with reference to FIGS. 3 and 4.

  As shown in FIG. 3, the encoding device 10 according to the present embodiment includes a table selection unit 101 (selection unit), a code amount acquisition unit 102 (acquisition unit), a parameter determination unit 103 (determination unit), and a quantum. A coding unit 104, a coding unit 105, a code amount control unit 106, and a storage unit 107 are included.

  The table selection unit 101 receives an input signal from the outside, and for each predetermined processing unit, from among one or more variable length coding tables stored in advance in the storage unit 107 based on the characteristics of the input signal, It is a functional unit that selects a variable-length coding table of 1. Here, the characteristics of the input signal may be, for example, the sum of absolute values or the variance of the signal values of the input signal. The variable-length coding table is a table for converting (variable-length coding) the input signal into a known code such as, for example, a Golomb Rice code or an Exponential Golomb code. The table selection unit 101 is a variable length coding table in which the amount of code when the variable length coding is performed on the input signal is reduced from one or more variable length coding tables stored in advance in the storage unit 107. select. For example, in the case of a variable length coding table of Golomb rice code, when the variance of the input signal is small, table selection section 101 is a variable length code having a short suffix length which is a parameter for specifying a variable length coding table of Golomb rice code. By selecting the coding table, it is possible to reduce the amount of code when coding with the variable length coding table. On the other hand, when the variance of the input signal is large, table selection section 101 selects a variable-length coding table with a long suffix length, thereby reducing the amount of code in the case of coding with the variable-length coding table. I can do it.

  The table selection unit 101 sends information (hereinafter sometimes referred to as a table index) indicating the selected variable length coding table to each of the code amount acquisition unit 102 and the coding unit 105. For example, in the case of a variable-length coding table of Golomb rice code, the suffix length may be used as a table index. The table selection unit 101 is realized by the encoding processor 502 shown in FIG.

  Note that the correspondence between the characteristic (absolute value sum or variance) of the input signal and the variable-length coding table is determined in advance, and information (for example, a formula or a look-up table) defining this correspondence is stored in the storage unit 107. The table selection unit 101 may select the variable length coding table by referring to the information defining the correspondence. For example, in the case of Golomb Rice code, since the sum of absolute values of input signals and the suffix length are in a logarithmic relationship, the table selection unit 101 calculates a suffix length from the sum of absolute values as information defining the correspondence. The variable length coding table may be selected based on the suffix length calculated from the characteristic value (absolute value sum) using

  Furthermore, although one or more variable length coding tables are stored in advance in storage unit 107 in the above description, the present invention is not limited to this, and table selection unit 101 has characteristics of the input signal (absolute value It is also possible to create a variable length coding table which is classified based on the sum or variance and encoded into a specific code (for example, a Huffman code) based on the appearance frequency of the input signal classified into each group . As a result, it is possible to create a variable-length coding table optimal for the input signal, and highly efficient coding can be realized. Further, in the above description, the one or more variable length coding tables are stored in advance in the storage unit 107. However, the present invention is not limited to this, and one or more variable length coding processing methods are prepared in advance. You may leave it. For example, in the case of Golomb-Rice code, variable length coding can be performed by calculating each parameter of prefix length and suffix value from an input signal and configuring a variable length code based on these values. Thus, when using a variable-length coding method that can be defined as predetermined arithmetic processing, instead of storing the variable-length coding table in the storage unit 107 by preparing this arithmetic logic in the coding device It can be done.

  Further, in the above, although the sum of absolute values and the variance are taken as examples of the characteristics of the input signal, the present invention is not limited to these, and an average value of the input signal may be used as the characteristics of the input signal. In general, when predictive coding is performed on speech or images, signals after prediction tend to be distributed around zero. However, when an input signal having a distribution other than 0 is input to the encoding device 10, selection of the variable-length coding table is performed based on the average value of the input signal to achieve higher efficiency. Encoding may be realized.

  In addition, although the table selection unit 101 selects one variable length coding table from the one or more variable length coding tables stored in the storage unit 107 based on the characteristics of the input signal, The variable length coding table stored in the storage unit 107 is not limited to a single table, and may be used. When the characteristics of the input signal hardly change, high efficiency coding may be performed even using a single variable length coding table. In this case, since it is not necessary to select one variable length coding table from among one or more variable length coding tables, the coding apparatus 10 need not have the table selection unit 101. In this case, the code amount acquisition unit 102 and the encoding unit 105 may perform the processing described later on the assumption that the single variable-length coding table is selected.

  The code amount acquisition unit 102 is a functional unit that receives a table index from the table selection unit 101 and acquires a reference code amount based on the table index. The method of acquiring the reference code amount by the code amount acquisition unit 102 will be described later. The code amount acquisition unit 102 sends the acquired reference code amount to the parameter determination unit 103. The code amount acquisition unit 102 is realized by the coding processor 502 shown in FIG.

  The parameter determination unit 103 is a functional unit that receives the reference code amount from the code amount acquisition unit 102, receives the allocated code amount from the code amount control unit 106, and determines the quantization parameter based on the reference code amount and the allocated code amount. is there. The determination method of the quantization parameter by the parameter determination unit 103 will be described later. The parameter determination unit 103 sends the determined quantization parameter to each of the quantization unit 104 and the encoding unit 105. The parameter determination unit 103 is realized by the encoding processor 502 shown in FIG.

  The quantization unit 104 is a functional unit that receives an input signal from the outside, receives a quantization parameter from the parameter determination unit 103, and quantizes the input signal with the quantization parameter to obtain a quantization value. For example, the quantization unit 104 performs quantization based on the following equation (1).

x1 = round (x0 / 2 ^q ) (1)

In the above equation (1), x0 represents an input signal, x1 represents a quantization value, and q represents a quantization parameter. Also, round () indicates a function that rounds off the decimal point. By performing the process using this function round (), when decoding the code encoded by the decoding device 20, it is possible to reduce the error of the decoded signal as compared to the case where rounding is not performed. Further, since the quantization value quantized by the above equation (1) becomes smaller according to the quantization width (2 ^{q of} equation (1)), the dispersion becomes smaller. The quantization unit 104 sends the quantization value obtained by the quantization to the coding unit 105. The quantization unit 104 is realized by the coding processor 502 shown in FIG.

  In addition, the quantization part 104 is not limited to performing quantization using said Formula (1), You may quantize based on another system. For example, H. The same processing as quantization in H.264 / MPEG-4 AVC (Advanced Video Coding) or the like may be performed.

  Encoding section 105 receives a table index from table selecting section 101, receives a quantization parameter from parameter determining section 103, receives a quantization value from quantization section 104, and performs variable-length coding on the quantization value. It is a functional unit. Specifically, encoding section 105 selects one variable-length coding table from among one or more variable-length coding tables stored in storage section 107 based on the table index and the quantization parameter. Then, variable length coding is performed on the quantization value by the variable length coding table. As described above, since the quantization value is smaller according to the quantization width and the dispersion is smaller, the coding unit 105 is a variable-length code corresponding to the quantization value whose dispersion is smaller than that of the input signal. By performing variable-length coding using a coding table, the amount of generated code can be reduced.

  Here, it is assumed that quantization is performed using the above-mentioned equation (1) and a variable-length coding table of Golomb's rice code is used. Further, as described above, the suffix length of the Golomb rice code is considered as a table index of the variable length coding table. The encoding unit 105 does not perform quantization on the variable-length coding table selected by the table selection unit 101 described above, that is, for example, in Equation (1), it is optimal when the quantization parameter q = 0. Since it is a variable-length coding table, if the received quantization parameter q is “0”, variable-length coding may be performed using the variable-length coding table selected by the table selection unit 101. Further, when the quantization parameter q determined by the parameter determination unit 103 is “1”, according to the above equation (1), since the quantization value is about half of the input signal, the dispersion is The variable-length coding table corresponding to the 1/2 signal (quantized value) is optimal. In the case of a Golomb-Rice code variable length coding table, this corresponds to a variable length coding table with a small suffix length. From this, assuming that the table index of the variable-length coding table selected by the table selection unit 101 is k, the encoding unit 105 stores the Golomb-Rise code stored in the storage unit 107 and having a suffix length k−q. Variable length coding may be performed using a variable length coding table. In addition, since the suffix length of the Golomb rice code is 0 or more, for example, when k−q is less than 0, the encoding unit 105 changes the variable length coding table of the Golomb Rice code whose suffix length is 0. It may be long coding. In addition, for example, when performing quantization by methods other than Formula (1), it is good also as what selects the variable-length coding table according to the change degree of the dispersion | distribution corresponding to each quantization parameter similarly to the above-mentioned.

  When variable length coding is performed using a variable length coding table of the Huffman code, the variable length coding table may be selected and coded by the same method. In the case of using a variable-length coding table of the Huffman code, a table corresponding to a plurality of characteristic values (sum or variance of absolute values of signals) is prepared in advance, and the table selection unit 101 selects one. As in the case of using the variable length coding table of Golomb's rice code, coding section 105 distributes (or obtains the absolute value of) the quantized value based on the table index of the variable length coding table of the Huffman code and the quantization parameter. The variable value coding may be performed on the quantized value using the sum (or sum) and using the variable length coding table corresponding to or closest to the sum. The variable length coding table of the Huffman code used here may be prepared in advance based on the appearance frequency of the signal classified according to the variance or the sum of absolute values.

  Also, for example, when performing variable-length coding with a variable-length coding table assuming an unsigned integer such as a Golomb-Rice code, and when the quantization value is a signed integer, the coding unit 105 According to Equation (2), variable length coding may be performed after converting the quantized value into an unsigned integer.

x2 = −2 · x1 (if x1 ≦ 0)
x2 = 2 · x1 + 1 (if x1> 0) (2)

  In the above equation (2), x1 represents a signed quantization value, and x2 represents a value converted to an unsigned integer. Note that, instead of performing conversion according to the above equation (2), code bits may be separately encoded, and variable length coding may be performed on the absolute value of the quantization value using a variable length coding table of Golomb's Rice code.

  The encoding unit 105 further encodes the table index and the quantization parameter to be used for the decoding process by the decoding device 20. In this case, the table index and the quantization parameter may be encoded with a fixed length according to, for example, a predetermined number of bits, or may be subjected to variable length coding using a predetermined variable length coding table. It is also good. Also, the table index and the quantization parameter may perform predictive coding. In the case where the characteristics of the table index and the quantization parameter do not significantly change, these values are predicted using values encoded in the past (for example, values encoded immediately before) (for example, 0th order prediction or 1st order prediction, etc. It is possible to reduce the amount of code generated by encoding the prediction residual.

  As described above, the encoding unit 105 encodes the table index, the quantization parameter, and the quantization value, and outputs the result as a code stream to the outside (for example, the network 40 or the like). Also, the coding unit 105 sends the code amount (generated code amount) of the coded code to the code amount control unit 106. The encoding unit 105 is realized by the encoding processor 502 shown in FIG.

  The code amount control unit 106 is a functional unit that receives the generated code amount from the encoding unit 105 and sets the allocated code amount. The code amount control unit 106 models the amount of code occupied in the reception buffer 30 (hereinafter referred to as “occupied code amount”), and the generated code amount in the coding unit 105 such that buffer underflow does not occur in the reception buffer 30. The allocation code amount is set so that

  For example, the code amount control unit 106 may use VBV (Video Buffer Verifier) in MPEG-2, or M.I. By assigning the same or similar modeling as a buffer model such as HRD (Hypothetical Reference Decoder) in H.264 / MPEG-4 AVC, the allocation code amount at which buffer underflow does not occur in the reception buffer 30 may be set.

  Specifically, for example, modeling of the occupied code amount of the reception buffer 30 as shown in FIG. 4 may be performed. In FIG. 4, the vertical axis indicates the code amount (occupied code amount) occupied by the reception buffer 30, and the horizontal axis indicates time. As shown in FIG. 4, the reception buffer 30 is in an initial state in which the occupied code amount is “0”, from which the code output from the encoding device 10 at a constant bit rate is accumulated. Thereafter, when a predetermined initial delay time has elapsed, the decoding device 20 starts decoding, and the code corresponding to the code amount in the first processing unit is removed from the reception buffer 30 by the decoding device 20. Thereafter, accumulation of the code output from the encoding device 10 and withdrawal of the code by the decoding device 20 are repeated for each predetermined processing unit. In the model shown in FIG. 4, buffer underflow occurs when the occupied code amount falls below “0”. That is, in each processing unit, when the code amount (generated code amount) of the code output by the encoding unit 105 is larger than the occupied code amount at the timing of removal of the code by the decoding device 20, buffer underflow occurs. .

  Therefore, the code amount control unit 106 sets a code amount equal to or smaller than the occupied code amount at the removal timing of the code in each processing unit as the allocated code amount. The encoding apparatus 10 can suppress the occurrence of buffer underflow by performing variable-length coding so that the quantization value of the input signal of a certain processing unit becomes a code amount smaller than the allocated code amount. It becomes. Here, the allocated code amount may be set as the same code amount as the occupied code amount at the removal timing of the code in each processing unit, but it is not limited to this, and the code amount smaller than this It may be set as

  Also, if the amount of occupied code of the reception buffer 30 is equal to or more than a certain amount, that is, if there is a margin in the amount of code that can be generated by the encoding unit 105, this margin is consumed at once in a certain processing unit It is easier to stabilize the generated code amount if this margin is divided by a plurality of subsequent processing units. From this, it is possible to stabilize the quality of the signal decoded by the decoding device 20 by setting the code amount smaller than the occupied code amount as the allocated code amount, particularly when the occupied code amount is large. .

  The code amount control unit 106 sends the set allocation code amount to the parameter determination unit 103. The code amount control unit 106 is realized by the coding processor 502 shown in FIG.

  The storage unit 107 is a functional unit that stores one or more variable length coding tables or the like that the table selection unit 101 and the coding unit 105 refer to. The storage unit 107 is realized by the auxiliary storage device 505 shown in FIG.

  The table selection unit 101, the code amount acquisition unit 102, the parameter determination unit 103, the quantization unit 104, the coding unit 105, the code amount control unit 106, and the storage unit 107 shown in FIG. It is shown and it is not limited to such composition. For example, a plurality of functional units illustrated as independent functional units in FIG. 3 may be configured as one functional unit. On the other hand, the function possessed by one functional unit in FIG. 3 may be divided into a plurality of parts and configured as a plurality of functional parts.

  Also, the table selection unit 101, the code amount acquisition unit 102, the parameter determination unit 103, the quantization unit 104, the coding unit 105, and the code amount control unit 106 shown in FIG. 3 are all the coding processor shown in FIG. Although realized by 502, it is not limited to this. That is, a program stored in a storage medium such as a storage device (for example, the auxiliary storage device 505 shown in FIG. 2) of at least one of the functional units described above is executed by a CPU (Central Processing Unit) It may be realized by

  FIG. 5 is a diagram showing an example of a graph showing the worst code amount for each quantization parameter. The details of the method of acquiring the reference code amount by the code amount acquisition unit 102 and the method of determining the quantization parameter by the parameter determination unit 103 will be described with reference to FIGS. 3 and 5.

  As described above, the table selection unit 101 is a variable-length code that reduces the code amount when variable-length coding is performed on the input signal based on the absolute value sum or variance of the input signal received from the outside. Table is selected. Conversely, given a variable-length coding table indicated by a certain table index, the possible values of the corresponding input signal will be limited to a certain finite range. For example, when a variable-length coding table is a table corresponding to the sum of absolute values of input signals ≦ 128, if this variable-length coding table is given, the values that can be taken by the input signal in the processing unit are the absolute values. It is limited to a group of arbitrary numbers (in the case where the input signal is an integer value, any integer) having a value sum of 128 or less. If the input signal is limited, the generated code amount when encoding these input signals with the corresponding variable-length coding table can be obtained. For example, this variable-length coding is performed by a method such as round robin. The table can obtain the maximum code amount when encoding the input signal. The obtained code amount is the maximum code amount when quantization is not performed, but the maximum code amount when encoding is similarly obtained for the quantized value for which quantization is performed. is there. This maximum code amount may be obtained in advance.

  When the quantization parameter and the quantization width have a monotonically increasing relationship, the larger the quantization parameter, the smaller the code amount at the time of encoding the quantization value quantized by the quantization parameter. For example, when the quantization width is expressed as an exponential function of the quantization parameter as in the above-mentioned equation (1), the encoded code amount can be approximated by a linear expression for the quantization parameter. This property is the same for the above-mentioned maximum code amount. Therefore, as shown in FIG. 5, modeling of the maximum code amount is performed in advance by a linear expression. Specifically, when encoding is performed using a variable-length coding table indicated by a certain table index, the maximum code amount in the case of using an arbitrary quantization parameter is obtained (for example, by round robin), as shown in FIG. As shown, a linear expression that exceeds all of these maximum code amounts is set (Equation (3) below).

WB = C ₁ xq + C ₂ (C ₁ <0, C ₂ > 0) (3)

In the above equation (3), C ₁ is the slope of the linear expression, and is a negative value because the code amount decreases with the increase of the quantization parameter q. C ₂ is an intercept of a linear expression. Further, WB indicates a code amount calculated by a linear expression, and hereinafter, it is referred to as a worst code amount. That is, Equation (3) is a linear expression indicating the worst code amount for each quantization parameter.

Then, for each table index, the inclination C ₁ and the intercept C ₂ of the linear expression shown in the above-mentioned equation (3) are obtained in advance. That is, if the table index is determined, a linear expression indicating the worst code amount is determined. This makes it possible to easily calculate the worst code amount for a specific quantization parameter for each table index. Further, information on the correspondence between the table index and the linear expression (i.e., the slope C ₁ and the intercept C ₂ ) indicating the worst code amount may be stored in the storage unit 107 in advance, for example. Then, code amount obtaining unit 102 refers to the storage unit 107 to identify the primary expression showing the worst amount of codes corresponding to the table index received from the table selection unit 101, for example, sections C ₂ of the linear expression , As a reference code amount. Incidentally, may be common between the table index for the slope C ₁ of the primary type, this case may be previously stored only sections C ₂ each table index.

  The parameter determination unit 103 specifies a linear expression indicating the worst code amount having the reference code amount acquired by the code amount acquisition unit 102, and the worst code amount does not exceed the allocated code amount received from the code amount control unit 106. Determine the quantization parameter. This makes it possible to suppress the occurrence of buffer underflow in the reception buffer 30. Specifically, since the worst code amount WB is modeled by the above equation (3), the following equation (4) may be satisfied for the assigned code amount B.

B C C ₁ x q + C ₂ (4)

  Further, when the quantization parameter q is arranged from the equation (4), the following equation (5) is obtained.

q ≧ (B−C ₂ ) / C ₁ (5)

  The parameter determination unit 103 may determine the quantization parameter q satisfying the equation (5), for example, determines the minimum quantization parameter q satisfying the equation (5), and sends it to the quantization unit 104. In order to stabilize the quality of the decoded signal of the decoding device 20, in order to suppress the fluctuation of the quantization parameter q, the value similar to the immediately preceding quantization parameter is quantized within the range satisfying the above equation (5). It may be determined as a conversion parameter.

  Note that, as shown in FIG. 5, the linear expression indicating the worst code amount shown in the above-mentioned equation (3) is set to a linear expression that exceeds all of the maximum code amounts corresponding to the respective quantization parameters. Not limited to this. For example, it is a graph connecting two points of the maximum code amount corresponding to each quantization parameter, and is set as a linear expression representing a graph in which the sum of absolute values of other maximum code amounts and differences is minimized. It is also good. In addition, the linear expression may be a linear expression showing a graph in which the intercept is shifted within a predetermined range (for example, a range of 1 to 2 times the intercept of the original graph) for the graph indicated by the linear expression set as described above. .

  FIG. 6 is a flowchart showing an example of the encoding process of the encoding apparatus according to the first embodiment. The flow of the encoding process of the encoding device 10 according to the present embodiment will be generally described with reference to FIG.

<Step S11>
The table selection unit 101 receives an input signal from the outside. Then, the process proceeds to step S12.

<Step S12>
The table selection unit 101 performs, for each predetermined processing unit, one or more variables stored in advance in the storage unit 107 based on the characteristics of the input signal (for example, the sum or variance of absolute values of signal values of the input signal). Among the long coding tables, one variable length coding table is selected. The table selection unit 101 sends information (hereinafter sometimes referred to as a table index) indicating the selected variable length coding table to each of the code amount acquisition unit 102 and the coding unit 105. Then, the process proceeds to step S13.

<Step S13>
The code amount acquisition unit 102 receives a table index from the table selection unit 101, and acquires a reference code amount based on the table index. Specifically, the code amount acquisition unit 102 refers to the storage unit 107 to specify a linear expression indicating the worst code amount corresponding to the table index received from the table selection unit 101, for example, an intercept of the linear expression the C _2, to obtain as the reference code amount. The code amount acquisition unit 102 sends the acquired reference code amount to the parameter determination unit 103. Then, the process proceeds to step S14.

<Step S14>
The parameter determination unit 103 receives the reference code amount from the code amount acquisition unit 102, receives the allocated code amount from the code amount control unit 106, and determines the quantization parameter based on the reference code amount and the allocated code amount. Specifically, the parameter determination unit 103 specifies a linear expression indicating the worst code amount having the reference code amount acquired by the code amount acquisition unit 102, and the assignment code received from the code amount control unit 106 for the worst code amount. Determine the quantization parameter so as not to exceed the amount. The parameter determination unit 103 sends the determined quantization parameter to each of the quantization unit 104 and the encoding unit 105. Then, the process proceeds to step S15.

<Step S15>
The quantization unit 104 receives an input signal from the outside, receives a quantization parameter from the parameter determination unit 103, quantizes the input signal with the quantization parameter, and obtains a quantization value. The quantization unit 104 sends the quantization value obtained by the quantization to the coding unit 105. Then, the process proceeds to step S16.

<Step S16>
Encoding section 105 receives a table index from table selecting section 101, receives a quantization parameter from parameter determining section 103, receives a quantization value from quantization section 104, and performs variable-length coding on the quantization value. . For example, assuming that the table index of the variable length coding table selected by the table selection unit 101 is k, the coding unit 105 stores the variable length of the Golomb Rice code stored in the storage unit 107 and having a suffix length k−q. The variable length coding is performed by the coding table. Then, the process proceeds to step S17.

<Step S17>
The code amount control unit 106 receives the generated code amount from the encoding unit 105, and sets an allocated code amount. The code amount control unit 106 models the occupied code amount occupied in the reception buffer 30, and allocates a code amount so as to be a generated code amount in the encoding unit 105 such that no buffer underflow occurs in the reception buffer 30. Set

  As shown in the above steps S11 to S17, the encoding device 10 performs the encoding process on the input signal input.

  As described above, in the encoding device 10 according to the present embodiment, a quantization parameter which is a code amount equal to or less than an allocated code amount by modeling the worst code amount for each table index indicating a variable-length coding table. Can be obtained with a small amount of processing to perform encoding. As a result, it is possible to realize signal encoding that suppresses the occurrence of buffer underflow in the reception buffer 30 with a small amount of processing. In addition, since the occurrence of buffer underflow is suppressed, it is not necessary to include means for avoiding buffer underflow such as skip in the encoding process, and the local SNR (Signal-to-Noise Ratio) is thereby reduced. It can be suppressed.

  In addition to the quantization, the encoding unit 105 also encodes the table index and the quantization parameter as described above. In this case, the parameter determination unit 103 may determine the quantization parameter in consideration of these code amounts. Specifically, for example, when encoding these parameters with a fixed number of bits, the number of bits is added to the right side in the above equation (4) for each processing unit, and then calculation of the quantization parameter is performed. You may go. Alternatively, for example, in the case of predictive encoding of the table index and the quantization parameter, the code amount at each value of the table index and the quantization parameter may be calculated for each processing unit and reflected in the above equation (4). Alternatively, more simply, the maximum code amount that can be generated in the case of predictive coding may be obtained in advance, and this maximum code amount may be added to calculate the quantization parameter.

  In addition, the encoding device 10 according to the present embodiment performs encoding using a variable-length encoding table corresponding to a signal with a smaller variance as the quantization parameter becomes larger. However, when the quantization parameter is equal to or more than a predetermined value, it may be difficult to reduce the code even if the quantization parameter is increased. For example, in the case of using Golomb Rice code, when the quantization parameter exceeds the suffix length as a table index of the variable length coding table selected by the table selection unit 101, a table with a smaller distribution is not prepared. By this, it becomes difficult to reduce the code even if the quantization parameter becomes large. Therefore, the variable-length coding table (that is, the variable-length coding table used when quantization is not performed) selected by the table selection unit 101 is one of a plurality of predetermined variable-length coding tables. It may be limited to a variable-length coding table corresponding to a signal with a large variance. For example, in the case of using a variable-length coding table of Golomb's rice code, the table selection unit 101 selects only a variable-length coding table whose suffix length is larger than a predetermined value determined in advance. In this case, the table selection unit 101 selects a suffix length k larger than the predetermined value k0. When the quantization parameter is q, the encoding unit 105 performs encoding using a variable-length encoding table with a suffix length k-q, but it does not matter if k-q is less than k0. When the variable length coding table used by the coding unit 105 for coding is not prepared by such selection of the variable length coding table by the table selection unit 101 (that is, k−q <0). And the worst code amount can be modeled more accurately.

  Further, in modeling of the worst code amount shown in the above equation (3), a linear expression which exceeds all the maximum code amounts corresponding to each quantization parameter is set, but the worst code amount is not necessarily linear. Sometimes it can not be modeled accurately. In such a case, the maximum code amount corresponding to a part of the quantization parameters may be regarded as an outlier, and modeling may be performed using a linear expression that falls below this. In this case, if quantization parameters corresponding to the maximum code amount regarded as outliers are used, buffer underflow may occur in the reception buffer 30, but other parts are accurately modeled. If so, it is possible to improve the coding efficiency as a whole. In this case, since the occurrence of the buffer underflow can not be completely avoided, it is necessary to perform processing such as skipping the subsequent encoding when the buffer underflow occurs. Here, in the portion regarded as the outlier of the above-mentioned worst code amount model, if the maximum code amount overrun of the model is sufficiently small, the number of insufficient bits at the occurrence of buffer underflow is sufficiently small (for example, , And several bits) can be expected. For example, when the encoding unit 105 performs encoding, first, the quantization parameter and the table index are encoded, and then the quantization value is encoded. Since this corresponds to the last few bits of the processing unit, even if buffer underflow occurs, the influence can be limited to a small part (for example, the last few symbols) in the processing unit. Furthermore, as for the encoding method of the quantization value, if the encoding information or information of the upper bit part of each quantization value in the processing unit is first encoded, and then the information of the lower bits is encoded, buffer underflow Since the influence range of the signal is in the lower bit part, complete loss of the signal can be prevented and decoding can be performed to a certain extent. For example, in the case of using a variable-length coding table of Golomb rice code, coding may be performed in the order of code information of a signal, a prefix part of the code, and a suffix part. In this case, since the range of the absolute value of the decoded signal is limited based on the length of the prefix portion of the code, decoding can be performed with a certain degree of accuracy from this information. Such an encoding method can prevent a significant reduction in SNR when buffer underflow occurs.

  In addition, when a buffer underflow occurs, the encoding unit 105 may perform encoding to the code amount at which the buffer underflow just occurs in the reception buffer 30, and may discard the subsequent codes in the processing unit. After that, the encoding unit 105 resumes encoding from the next processing unit. The decoding device 20 reads all the codes in the reception buffer 30, and decodes up to the part that can be decoded in the processing unit. For signals that can not be decoded, replacement may be performed with a predetermined value (for example, “0” or the like). After that, the decoding device 20 can read the code from the top of the reception buffer 30 in the next processing unit, and resume the decoding.

  In addition, although the method of suppressing the occurrence of buffer underflow in the reception buffer 30 has been described for the encoding by the encoding unit 105 described above, the occurrence of buffer overflow is also suppressed by the same or similar method. Is possible. Specifically, the code amount control unit 106 outputs a minimum code amount at which a buffer overflow does not occur, instead of the allocated code amount. Instead of the worst code amount, the coding apparatus 10 models in advance the code amount generated at the minimum in each variable length coding table for each processing unit. The parameter determination unit 103 sets a quantization parameter based on the table index and the code amount output from the code amount control unit 106 based on this model. In order to suppress the occurrence of buffer overflow, the upper limit value of the quantization parameter can be obtained because it suffices to satisfy the mathematical expression in the inverted form of the inequality sign of the above equation (4), and any quantization can be obtained within this range. The parameters should be selected. Note that the upper limit value of the quantization parameter may fall below a predetermined minimum value (for example, “0” when performing quantization in the above equation (1)), but this is the minimum value It means that even if buffer overflow remains possible. If a buffer overflow is likely to occur due to a shortage of actually generated code amount, a surplus code is added to the end of the processing unit code to add up the generated code amount and suppress the occurrence of the buffer overflow. Good (so-called bit stuffing process).

  Also, the method of suppressing the occurrence of buffer overflow described above may be used in combination with the method of suppressing the occurrence of buffer underflow described above. In this case, the parameter determination unit 103 may calculate the upper limit value and the lower limit value of the quantization parameter from the allocated code amount, the minimum code amount, and the table index, and set the quantization parameter within this range. .

  Further, in the encoding device 10 according to the present embodiment, the table index and the quantization parameter are determined for each processing unit, but the present invention is not limited to this. For example, the quantization parameter may be determined for each first processing unit, and the table index may be selected for each second processing unit smaller than the first processing unit. For example, when transform coding is performed, the variance of the code may be different for each position of transform coefficients. In such a case, for example, different table indexes are selected by the first half coefficient (for example, a signal having a low spatial frequency in the case of an image) and the second half coefficient (for example, a signal having a high spatial frequency in the case of an image) It is possible to improve the coding efficiency. In this case, the parameter determination unit 103 models the worst code amount for each second processing unit in the first processing unit, and the sum of the worst code amounts corresponding to the quantization parameter is less than the allocated code amount. The quantization parameter may be set to.

Second Embodiment
An encoding apparatus according to the second embodiment will be described focusing on differences from the encoding apparatus 10 according to the first embodiment. In the first embodiment, the operation of performing variable-length coding by selecting a variable-length coding table for each predetermined processing unit has been described. In the present embodiment, an operation of reusing information of the previously selected variable length coding table will be described. The overall configuration of the transmission system according to the present embodiment and the hardware configuration of the encoding apparatus are the same as the configuration described in the first embodiment.

  FIG. 7 is a diagram showing an example of a block configuration of the coding apparatus according to the second embodiment. The block configuration and operation of the encoding device 10a according to the present embodiment will be described with reference to FIG.

  As shown in FIG. 7, the encoding device 10a according to the present embodiment includes a table selection unit 101a (selection unit), a code amount acquisition unit 102 (acquisition unit), a parameter determination unit 103a (determination unit), and a quantum. A coding unit 104, a coding unit 105a, a code amount control unit 106, and a storage unit 107 are included. The operations of the code amount acquisition unit 102, the quantization unit 104, the code amount control unit 106, and the storage unit 107 are the same as the contents described in the first embodiment.

  The table selection unit 101a receives an input signal from the outside, and for each predetermined processing unit, from among one or more variable length coding tables stored in advance in the storage unit 107, based on the characteristics of the input signal, It is a functional unit that selects a variable-length coding table of 1. The method of selecting the variable-length coding table based on the characteristics of the input signal may be the same as or similar to the operation of the table selection unit 101 of the first embodiment. The table selection unit 101a further generates selection information. Here, the selection information indicates which variable-length coding table is to be used among the variable-length coding table selected in the processing unit and the variable-length coding table previously selected (for example, immediately before). It is information.

  If the characteristics (such as variance or sum of absolute values) of the input signal are the same or similar as before, it is necessary to encode the table index in the encoding unit 105a by using the previously selected variable length encoding table. Can be eliminated to reduce the amount of code. On the other hand, when the characteristic of the input signal is different from before, the table selection unit 101a can improve the coding efficiency by newly selecting a variable-length coding table.

  Further, the setting of the selection information by the table selection unit 101a is performed, for example, as follows. First, in each processing unit, the table selection unit 101a acquires a characteristic value (such as a variance value or a sum of absolute values) of an input signal in order to select a variable-length coding table. The table selection unit 101a compares the acquired characteristic value with the characteristic value corresponding to the previously selected variable length coding table, and if the difference is less than a predetermined threshold, the previously selected variable length coding is performed. Select the table again. If the difference is larger than the predetermined threshold, the table selection unit 101a newly selects a variable-length coding table based on the characteristic value of the input signal in the processing unit. In the process of comparing the characteristic values, instead of comparing the characteristic values themselves, threshold determination may be performed on the difference between their logarithmic values. Generally, since the generated code amount at the time of encoding is correlated with the logarithm of the characteristic value, it is possible to select a variable-length encoding table with high encoding efficiency by performing such comparison.

  The table selection unit 101a sends a table index indicating the selected variable length coding table to the code amount acquisition unit 102 and the coding unit 105a, and the generated selection information to the parameter determination unit 103a and the coding unit 105a. send. The table selection unit 101a is realized by the encoding processor 502 shown in FIG.

  The parameter determination unit 103a is a functional unit that receives the reference code amount from the code amount acquisition unit 102, receives the allocated code amount from the code amount control unit 106, receives selection information from the table selection unit 101a, and determines the quantization parameter. . The parameter determination unit 103a is realized by the coding processor 502 shown in FIG. The method of determining the quantization parameter by the parameter determination unit 103a is the operation of the parameter determination unit 103 of the first embodiment when the selection information indicates that the variable length coding table is newly selected by the table selection unit 101a. The quantization parameter may be determined by the same or similar method. On the other hand, when the selection information indicates that the variable-length coding table previously selected by the table selection unit 101a is selected, the parameter determination unit 103a determines the quantization parameter by the following operation.

  In the first embodiment, the parameter determination unit 103 specifies a linear expression indicating the worst code amount shown in the above equation (3), which has the reference code amount acquired by the code amount acquisition unit 102. If the selection information indicates that the variable length coding table previously selected by the table selection unit 101a is selected, the parameter determination unit 103a uses the variable length coding table previously selected by the table selection unit 101a. . In this case, the code amount may be larger than when coding is performed using a variable-length coding table (optimum variable-length coding table) newly selected by the table selection unit 101a. In this case, using the linear expression indicating the worst code amount shown in the above equation (3), the prediction accuracy of the worst code amount is deteriorated, and as a result, there is a problem that the coding efficiency is reduced. Therefore, as shown in the following equation (6), the worst code amount is modeled by adding an increase of the code amount due to the difference of the variable length coding table as another term.

_{WB = C 1 xq + C 2} + dif [t 1, t 2] (C 1 <0, C 2> 0) ··· (6)

In the above equation (6), C ₁ and C ₂ may be the same as or similar to the corresponding values in the above equation (3). t ₁ indicates a table index of a variable-length coding table corresponding to the characteristic value (absolute value sum or variance etc.) of the input signal, and t ₂ indicates a variable-length coding table previously selected by the table selection unit 101 a. Indicates a table index. Then, dif [t ₁ , t ₂ ] indicates the difference between the encoded amount when the table index is t ₁ and when the table index is t ₂ . dif [t ₁ , t ₂ ] may be “0” when t ₁ = t ₂ , and becomes larger as the variances of the variable-length coding tables of t ₁ and t ₂ differ. The value of dif [t ₁ , t ₂ ] can be obtained by coding the input signal corresponding to an arbitrary table index t ₁ with a variable length coding table corresponding to the table index t ₂ , for example, by round-robin coding etc. It can be acquired. The coding is performed round-round for each t ₁ and t ₂ to obtain the worst code amount, and the difference from the worst code amount in the case of t ₁ = t ₂ is taken as dif [t ₁ , t ₂ ] Become. If the worst code amount is modeled according to the above equation (6), the method of determining the quantization parameter by the parameter determining unit 103a thereafter is similar to the operation of the parameter determining unit 103 of the first embodiment. Specifically, the following equation (7) may be satisfied for the allocation code amount B.

B C C ₁ x q + C ₂ + dif [t ₁ , t ₂ ] (7)

  Further, when the quantization parameter q is arranged from the equation (7), the following equation (8) is obtained.

q ((B-C _2- dif [t ₁ , t ₂ ]) / C ₁ (8)

  The parameter determination unit 103a may determine the quantization parameter q that satisfies the equation (8), for example, determines the smallest quantization parameter q that satisfies the equation (8), and sends it to the quantization unit 104.

As described above, although the difference between the worst code amounts when the table index is different is modeled as the difference dif [t ₁ , t ₂ ] for the two-dimensional table, a variable-length coding table prepared in advance is prepared. It is also possible to perform modeling with a one-dimensional table, depending on the characteristics of. For example, when using a variable-length coding table of Golomb's rice code, a certain variable-length coding table (for example, suffix length = k) and its adjacent variable-length coding table (for example, suffix length = k + 1, k-1) And corresponds to an input signal whose sum of absolute values is twice or half. This relationship holds for any variable-length coding table with a suffix length of 2 or more. From this, the difference in the worst code amount when using different variable-length coding tables mostly depends on the absolute value of each suffix length It will depend heavily on the difference. From this, by performing modeling as shown in the following equation (9) instead of the equation (7), it is possible to reduce the amount of information held in the encoding device 10a.

B C C ₁ x q + C ₂ + dif [t _1- t ₂ ] (9)

  Further, even in the case of using a variable-length coding table of a code other than Golomb's rice code, the same modeling as the above-mentioned equation (9) can be performed depending on the prepared variable-length coding table. For example, when preparing a variable-length coding table based on the Huffman code, the probability distribution serving as the source of the variable-length coding table may be a distribution in which the table index is changed at a predetermined ratio. Specifically, for example, if a normal distribution is prepared such that the variance is doubled each time the table index increases by “1”, and the variable length coding table of the Huffman code is configured based on this distribution, the formula ( It becomes possible to perform modeling like 9).

  The parameter determination unit 103a sends the determined quantization parameter to each of the quantization unit 104 and the encoding unit 105a.

  The encoding unit 105a receives the table index and the selection information from the table selection unit 101a, receives the quantization parameter from the parameter determination unit 103a, receives the quantization value from the quantization unit 104, and performs variable-length coding on the quantization value. Is a functional unit that Specifically, when the selection information indicates that the variable-length coding table previously selected by the table selection unit 101a is selected, the coding unit 105a is a variable encoded previously stored therein. The long coding table is used to perform variable length coding on the quantization value. The coding unit 105 a is a table index (a variable length coding table previously selected by the table selection unit 101 a), as in the coding unit 105 of the first embodiment, for the variable length coding table to be actually used. It is good also as what is selected based on a table index which shows, and a quantization parameter. On the other hand, when the selection information indicates that the variable-length coding table has been newly selected by the table selection unit 101a, the coding unit 105a is variable by the same or similar operation as the coding unit 105 of the first embodiment. Long encoding may be performed. The encoding unit 105a internally holds table index information so that it can be used in the subsequent processing unit. The encoding unit 105a is realized by the encoding processor 502 shown in FIG.

  The encoding unit 105a performs encoding of the table index by a method different from the same method as the encoding unit 105 of the first embodiment. Specifically, the encoding unit 105a first encodes selection information (for example, a flag), and then, changes the variable only when the selection information selects a variable-length coding table selected in the processing unit. Encode the table index of the long encoding table. This reduces the amount of code at the time of encoding the table index. In other operations, the encoding unit 105a performs the same or similar operation as the encoding unit 105 of the first embodiment. As described above, the encoding unit 105a encodes the selection information, the table index (if necessary), the quantization parameter and the quantization value, and outputs the result as a code stream to the outside (for example, the network 40 or the like).

  As described above, the encoding device 10a according to the present embodiment is a quantization parameter that suppresses the occurrence of buffer underflow in the reception buffer 30 with a small amount of processing, similarly to the encoding device 10 according to the first embodiment. And encoding can be performed. As a result, the amount of processing required for the encoding process can be reduced, and it is not necessary to include means for avoiding buffer underflow due to skipping or the like, and local degradation of the SNR of the decoded signal can be suppressed.

  While certain embodiments of the present invention have been described, these embodiments have been presented by way of example only, and are not intended to limit the scope of the invention. These novel embodiments can be implemented in various other forms, and various omissions, substitutions, and alterations can be made without departing from the scope of the invention. These embodiments are included in the scope and the gist of the invention, and are included in the invention described in the claims and the equivalents thereof.

Reference Signs List 1 transmission system 10, 10a encoding device 20 decoding device 30 reception buffer 40 network 101, 101a table selection unit 102 code amount acquisition unit 103, 103a parameter determination unit 104 quantization unit 105, 105a encoding unit 106 code amount control unit 107 Storage unit 501 Input I / F
502 Encoding Processor 502a Cache 503 RAM
504 Output I / F
505 Auxiliary storage device 506 bus

Claims (11)

A coding apparatus that performs variable-length coding on a quantized value obtained by quantizing an input signal based on a quantization parameter, comprising:
A selector configured to select one variable length coding table from among two or more variable length coding tables based on the characteristics of the input signal;
An acquisition unit for acquiring the maximum code amount that may occur for each of the quantization parameters according to the variable-length coding table selected by the selection unit;
A determination unit that determines the quantization parameter based on the maximum code amount for each quantization parameter and an allocated code amount determined by an occupied code amount of a reception buffer;
A quantization unit that quantizes the input signal to obtain the quantization value using the quantization parameter determined by the determination unit;
Based on the quantization parameter determined by the determination unit, one variable-length coding table is selected from the two or more variable-length coding tables, and the quantization value is determined by the variable-length coding table. An encoding unit that performs variable-length encoding on
An encoding device comprising: The acquisition unit acquires a function that determines the maximum code amount that may occur for each of the quantization parameters,
The determination unit determines the quantization parameter based on the maximum code amount for each quantization parameter obtained from the function and the allocated code amount obtained from the occupied code amount of the reception buffer. The encoding device according to 1.   The encoding apparatus according to claim 1, wherein the selection unit selects one variable-length coding table based on the variance or sum of absolute values of the input signal as the characteristic of the input signal.   A code for setting the allocation code amount such that the code amount of a code for which variable length coding is performed by the coding unit is smaller than the occupied code amount at the timing when the code of the reception buffer is removed by the decoding device The encoding device according to claim 1, further comprising a quantity control unit. The variable length coding table selected by the coding unit is a table based on Golomb rice code,
The coding unit corresponds to a variable length coding table corresponding to a smaller variance than a variance corresponding to the variable length coding table selected by the selection unit as the quantization parameter determined by the determination unit is larger. selected,
The encoding device according to claim 2 , wherein the function acquired by the acquisition unit is a linear function .   The encoding unit selects a variable length encoding table indicated by using a value obtained by subtracting the quantization parameter determined by the determination unit from the suffix length of the variable length encoding table selected by the selection unit as an index. The encoding device according to claim 5.   The encoding apparatus according to claim 6, wherein the selection unit selects a variable-length coding table having a suffix length larger than a predetermined value. The quantization width of the quantization by the quantization unit is an exponential function of the quantization parameter,
The function number, the encoding apparatus according to claim 2 represented by a linear equation of the quantization parameter. The selection unit may perform variable length coding previously selected when a difference between the value indicating the characteristic of the input signal and the value of the characteristic corresponding to the previously selected variable length coding table is less than a predetermined threshold. Select the table again,
If the variable length coding table previously selected is selected by the selection section, the coding section performs variable length coding on the quantized value according to the variable length coding table previously used for coding. The coding apparatus according to claim 1, wherein The acquisition unit is based on the maximum code amount corresponding to the reselected variable length coding table and the characteristics of the input signal when the variable length coding table previously selected by the selection unit is selected again. and it acquired the difference between the maximum code amount corresponding to the variable length coding table,
The determination unit determines the quantization parameter on the basis of the maximum code amount for each quantization parameter, the difference, and the allocated code amount determined by the occupied code amount of the reception buffer. The encoding device according to claim 1. A coding method of a coding apparatus for performing variable-length coding on a quantized value obtained by quantizing an input signal based on a quantization parameter, comprising:
A selection step of selecting one variable length coding table from among two or more variable length coding tables based on characteristics of the input signal;
An acquiring step of acquiring a maximum code amount that may be generated for each of the quantization parameters according to the selected variable length coding table;
A determination step of determining the quantization parameter based on the maximum code amount for each quantization parameter and an allocated code amount determined by an occupied code amount of the reception buffer;
Quantizing the input signal using the determined quantization parameter to obtain the quantization value;
Based on the determined quantization parameter, one variable-length coding table is selected from the two or more variable-length coding tables, and the variable-length coding table is used to change the variable length of the quantization value. An encoding step for performing encoding;
Encoding method.

Images