JP2007324971A - Encoder, encoding method, and program - Google Patents

Abstract

<P>PROBLEM TO BE SOLVED: To provide an encoder achieving encoding processing more efficiently. <P>SOLUTION: The encoder 2 combines a fixed encoding processing without depending on an inputted data file with a variable encoding processing depending on the inputted data file in series to reduce a buffering load. The fixed encoding processing is executed in the pre-stage of buffering processing, thus reducing the size of data to be buffered for reducing the load of the buffering. The variable encoding processing is performed to a buffered symbol additionally, thus achieving the same compression ratio as that of the variable encoding processing. <P>COPYRIGHT: (C)2008,JPO&INPIT

Description

  The present invention relates to an encoding method for encoding a data file.

For example, in Patent Document 1, the number of limited colors to be obtained is specified, a target number of limited colors is extracted from the color distribution of the color image, and the frequency distribution of the number of pixels assigned to each limited color is extracted from the color image. Count and use the frequency distribution to determine the probability of occurrence, assign a Huffman code to each of the limited colors, and for each pixel, if the limited color of the pixel is the same as the limited color of the adjacent pixel, a positional relationship code, otherwise A method is disclosed in which a code of a limited color is assigned and stored as compressed data, and finally, additional information of the compressed data is stored.
JP-A-6-274609

  The present invention has been made from the above-described background, and an object thereof is to provide an encoding device that realizes more efficient encoding processing.

[Encoding device]
In order to achieve the above object, an encoding apparatus according to the present invention encodes an input data file using an entropy encoding method that does not depend on unprocessed data, and generates an intermediate code. Generating means and second code generating means for generating an output code of the input data file based on the intermediate code generated by the first code generating means.

  Preferably, the second code generation means generates the output code using an encoding method depending on unprocessed data.

  Preferably, the second code generation means generates the output code based on statistical processing for the entire data file.

  Preferably, it further includes statistical processing means for performing the statistical processing in parallel with at least a part of the encoding processing by the first code generating means, and the second code generating means includes the statistical processing means. The output code is generated based on the result of the statistical processing according to the above.

  Preferably, the first code generation unit performs an encoding process in a larger data unit than the second code generation unit.

  Preferably, the first code generation means encodes the input data file using a pre-designed fixed code table, and the second code generation means responds to each of the input data files. The output code is generated using a variable code table designed in this way.

  Preferably, the apparatus further comprises storage means for storing the intermediate code generated by the first code generation means, and the second code generation means encodes the intermediate code stored in the storage means, The output code is generated.

[Encoding method]
In addition, the encoding method according to the present invention encodes an input data file using an entropy encoding method that does not depend on unprocessed data, generates an intermediate code, and based on the generated intermediate code The output code of the input data file is generated.

[program]
In addition, the program according to the present invention encodes an input data file using an entropy encoding method that does not depend on unprocessed data, generates an intermediate code, and based on the generated intermediate code And causing the computer to generate an output code of the input data file.

  According to the encoding apparatus of the present invention, more efficient encoding processing can be realized.

First, in order to help understanding of the present invention, its background and outline will be described.
The Huffman code is a widely used coding method because of its simple implementation configuration. Originally, an appearance frequency distribution of each input symbol is obtained for each input data file, and a code table (variable code table) corresponding to the distribution is designed (hereinafter referred to as a variable Huffman code).
However, in Huffman codes that are commonly used, including the international standard JPEG, a code table (fixed code table) designed for the frequency distribution assumed in advance is often applied fixedly (below) This is called the fixed Huffman code).

When the Huffman code is applied to image compression, a conversion process called source coding is performed in the previous stage to generate a symbol to be encoded. At this time, if the encoding process in the subsequent stage applies a fixed Huffman code, the compression rate decreases when the output (symbol) of the source coder does not match the expected appearance frequency.
Therefore, for example, there is a problem that it is difficult to use a biased source coder that is efficient for a CG image but relatively inefficient for a photographic image.
On the other hand, in the variable Huffman coding process, since the statistics of the symbols are taken in the first pass and coding is done in the second pass, the processing speed is slower than the case where the fixed Huffman code is applied. In addition, when the variable Huffman code is applied, it is necessary to buffer the entire input data for the two-pass processing, and processing time and memory cost become a bottleneck.

FIG. 1 is a diagram illustrating a functional configuration of an encoding program 9 to which a variable Huffman code is applied.
One of the causes for delaying the variable Huffman coding process is the buffering process. In particular, when a variable Huffman code is applied to image compression, symbols are often generated by performing source coding by the source coder 900, but it is necessary to store these symbols in the symbol buffer 910 separately from the image.
For example, when performing source coding using a predictive coding method, the source coder 900 converts a pixel value of input image data into a prediction error value, and uses the prediction error value as a symbol. This prediction error value is buffered in the symbol buffer 910 until the code table is created. That is, the symbol counting unit 920 counts the symbols output from the source coder 900, and the code table generating unit 930 generates a variable code table based on the counting result (appearance frequency) by the symbol counting unit 920, and the Huffman code The conversion unit 940 reads a symbol from the symbol buffer 910 on the condition that the code table is generated by the code table generation unit 930, and encodes the read symbol.
Since there are as many prediction error values as the number of pixels to be buffered, the symbol buffer 910 requires memory for one page of image, and writing and reading of this memory is also required for one page. In particular, in the case of predictive coding, the processing of the source coder 900 is light and parallelization is relatively easy, which increases the relative processing load of buffering.

  Therefore, in order to reduce the buffering load, the encoding device 2 according to the present embodiment is input with a fixed encoding process (for example, a fixed Huffman encoding process) that does not depend on the input data file. A variable encoding process depending on the data file (for example, a variable Huffman encoding process adapted to the input image) is combined in series. Among these, the fixed encoding process is executed before the buffering process, thereby reducing the data size of the data to be buffered and reducing the buffering load. For buffered symbols (codes encoded in the fixed encoding process), the same compression rate as the variable encoding process is achieved by additionally performing the variable encoding process. To do. In the following description, a code (buffered code) encoded by the fixed encoding process is referred to as an intermediate code.

The fixed encoding process preferably satisfies the following conditions.
[Condition 1]
Processing that does not depend on the global nature of the data file (image data in this example). More specifically, it is an encoding process that does not depend on at least unprocessed data (unencoded partial data). In other words, it is an encoding process having a dependency relationship with other data within the limit of local properties that can be processed in one pass.
[Condition 2]
It must be relatively fast compared to variable encoding.
[Condition 3]
Less overhead due to combining fixed encoding processing and variable encoding processing.

For example, regarding the condition 1, the statistical processing of the entire page necessary for generating the variable Huffman code is a global property. Conversely, context modeling is a local property because it does not depend on subsequent data (ie, raw data).
Condition 2 is required for the purpose of reducing the load of buffering, and the variable encoding process must be heavy by a process such as bit shift, so that the fixed encoding process should be light. . However, if the sum of the loads of the fixed encoding process and the variable encoding process is approximately the same as the processing load of the original entropy encoding, the fixed encoding process may be slightly heavier in balance. Absent.
As for condition 3, it is conceivable as a means for satisfying this condition to arrange the processing units. For example, when applying a variable Huffman code in a variable encoding process and applying an arithmetic code in a fixed encoding process, a decoding process for the arithmetic encoding is required before the variable encoding process. The overhead becomes too big.
By combining fixed encoding processing and variable encoding processing satisfying the above conditions, the amount of data to be buffered can be reduced, so the buffering processing time can be reduced and the memory size can also be reduced statistically. It becomes like this.

[Hardware configuration]
First, the hardware configuration of the encoding device 2 in the present embodiment will be described.
FIG. 2 is a diagram illustrating a hardware configuration of the encoding device 2 to which the encoding method according to the present invention is applied, centering on the control device 20.
As illustrated in FIG. 2, the encoding device 2 includes a control device 20 including a CPU 202 and a memory 204, a communication device 22, a recording device 24 such as an HDD / CD device, an LCD display device or a CRT display device, and a keyboard. A user interface device (UI device) 26 including a touch panel and the like is included.
The encoding device 2 is, for example, a processing device provided in the printer device 10, acquires image data via the communication device 22 or the recording device 24, and encodes the acquired image data.

[Encoding program]
FIG. 3 is a diagram illustrating a functional configuration of the encoding program 5 which is executed by the control device 20 (FIG. 2) and implements the encoding method according to the present invention.
As illustrated in FIG. 3, the encoding program 5 includes a source coder 500, a first encoding unit 510, a symbol counting unit 520, an intermediate code storage unit 530, a code table generation unit 540, and a second encoding unit 550.
The encoding program 5 is recorded on, for example, the recording medium 240 illustrated in FIG. 2, and is supplied to the control device 20 (FIG. 2) via the recording medium 240.
In this example, a case where image data is encoded will be described as a specific example.

In the encoding program 5, the source coder 500 generates a symbol to be subjected to entropy encoding processing based on the input data file, and outputs the generated symbol to the first encoding unit 510 and the symbol counting unit 520. To do.
The source coder 500 in this example predicts the pixel value of the target pixel using a predetermined prediction method, calculates a difference value between the predicted value and the target pixel value (hereinafter, prediction error value), and calculates the calculated prediction. The error value is a symbol. Note that the source coder 500 in this example counts the number of consecutive predictions (runs) when the calculated prediction error value is 0 (that is, when the prediction is correct), and is counted. The continuous number (the number of times that the prediction error value is continuously 0) is defined as a symbol.

The first encoding unit 510 encodes the symbol input from the source coder 500 using an entropy encoding method that does not depend on unprocessed data (for example, an encoding method using a pre-designed fixed code table). The generated code (that is, the intermediate code) is output to the intermediate code storage unit 530.
The first encoding unit 510 of this example encodes a symbol input from the source coder 500 with a fixed Huffman code to generate an intermediate code.

The symbol counting unit 520 counts the symbols input from the source coder 500, performs predetermined statistical processing based on the counting result, and outputs the statistical processing result to the code table generation unit 540. The symbol counting unit 520 is an example of a statistical processing unit according to the present invention, and performs statistical processing required for code table generation.
The symbol counting unit 520 of this example counts the appearance frequencies of all symbols input from the source coder 500, and generates the appearance frequency distribution of each symbol value.

The intermediate code storage unit 530 stores the intermediate code input from the first encoding unit 510 in a memory and supplies the stored intermediate code to the second encoding unit 550.
The intermediate code storage unit 530 of the present example buffers the intermediate code encoded by the first encoding unit 510 until at least the code table of the variable Huffman code is generated by the code table generation unit 540. The buffered intermediate code is output to the second encoding unit 550 on condition that the second encoding unit 550 can execute the variable encoding process.

The code table generation unit 540 generates a code table used for variable encoding processing based on the statistical processing result by the symbol counting unit 520 and outputs the generated code table to the second encoding unit 550. The code table generated by the code table generation unit 540 is designed according to each input data file and is called a variable code table.
The code table generation unit 540 of the present example generates a code table of variable Huffman codes according to the appearance frequency distribution of each symbol.

The second code unit 550 generates an output code of the input data file based on the variable code table input from the code table generation unit 540 and the intermediate code stored in the intermediate code storage unit 550.
The second encoding unit 550 of this example sequentially encodes the intermediate codes buffered in the intermediate code storage unit 530 using the code table of the variable Huffman code input from the code table generation unit 540, and outputs the result. Generate a code.

FIG. 4 is a diagram for explaining source coding by the source coder 500 (FIG. 3).
As illustrated in FIG. 4A, when generating the symbol of the target pixel X, the source coder 500 refers to any one of the pixels A to C upstream of the scanning direction (that is, the processed pixel). The prediction error value is calculated.
For example, the source coder 500 calculates a prediction error (symbol) using (Formula 1) illustrated in FIG. That is, the difference value between the pixel value of the target pixel X and the pixel value of the pixel A immediately before the main scanning direction is the prediction error value (symbol).
Further, the source coder 500 may calculate a prediction error (symbol) using (Equation 2) illustrated in FIG. In this case, a value obtained by subtracting the pixel value of the pixel C from the sum of the pixel values of the pixel A and the pixel B (pixels directly above in the sub-scanning direction) is the predicted value, and the difference between the predicted value and the pixel value of the target pixel X is Become a symbol.
Further, the source coder 500 may calculate the prediction error using (Equation 3) illustrated in FIG. In this case, the average pixel value of the pixel A and the pixel B becomes a predicted value, and the difference between this predicted value and the pixel value of the target pixel X becomes a symbol.

Furthermore, when the prediction error becomes 0, the source coder 500 counts the continuous number (run), and uses the continuous number (run) as a symbol. As a result, as illustrated in FIG. 4C, for the pixel group for which the prediction is correct, the predictive number (run) of the prediction is a symbol, and for the pixel group for which the prediction has failed, the prediction error value is Become a symbol.
When image data is source-coded, it can be expected that the appearance frequency distribution of prediction error values is biased to be close to zero.

5A illustrates the fixed Huffman code by the first encoding unit 510 (FIG. 3), and FIG. 5B illustrates the format of the intermediate code stored in the intermediate code storage unit 530 (FIG. 3). It is a figure illustrated.
The first encoding unit 510 of this example encodes a symbol using the fixed code table 512 illustrated in FIG.
That is, when a symbol is a continuous number (run), a bit string combining an identifier “00” and a continuous number expressed in 8 bits is an intermediate code. In this example, since the continuous number is expressed by 8 bits, the run is divided by the maximum run length 256.
When the symbol is a prediction error value between −8 and 7, a bit string obtained by combining the identifier “01” and the prediction error value represented by 4 bits (that is, the lower 4 bits of the prediction error value) is intermediate. Sign.
When the symbol is a prediction error value smaller than −8 or a prediction error value larger than 7, a bit string obtained by combining the identifier “10” and the prediction error value expressed in 8 bits is an intermediate code.
Although this intermediate code is very simple, it is a fixed Huffman code that assumes a high probability of occurrence of prediction error values from -8 to 7.

The fixed code table 512 is designed in advance and does not depend on input image data, and therefore satisfies the above condition 1. Note that the fixed code table 512 of this example does not minimize the code amount, but it is sufficient if the amount of data to be buffered is reduced, and it is not always necessary to minimize the code amount. Is suitable.
Further, the second encoding unit 550 of this example applies a variable Huffman code, and this fixed code table 512 is a fixed Huffman code designed in advance based on the characteristic that the prediction error value increases near 0, Since both are Huffman codes, the above condition 3 is satisfied.

The intermediate code storage unit 530 of this example manages the intermediate code in units of 4 bits in order to satisfy the above condition 2. That is, the intermediate code storage unit 530 of the present example buffers the intermediate code using the buffering code format 514 illustrated in FIG.
Specifically, since the intermediate code illustrated in FIG. 5A includes an identifier (2 bits) and an attribute value (4 bits or 8 bits), it cannot be managed in units of 4 bits. Therefore, the intermediate code storage unit 530 of this example integrates the four intermediate codes into one format. That is, as illustrated in FIG. 5B, the intermediate code storage unit 530 collects the identifiers (# 0 to # 3) of the four intermediate codes, and attribute values (# 0 to # 3) of the four intermediate codes. As a result, management in units of 4 bits is realized.
As a result, excessive bit shift processing is avoided and the condition 2 is easily satisfied.

[Overall operation]
Next, the overall operation of the encoding device 2 (encoding program 5) will be described.
FIG. 6 is a flowchart of the encoding process (S10) by the encoding program 5 (FIG. 3).
As shown in FIG. 6, in step 100 (S100), when image data is input from the outside, the encoding program 5 (FIG. 3) sets a target pixel X in the input image data in the scanning order. A target pixel is a pixel to be processed.
The source coder 500 (FIG. 3) calculates the difference between the pixel value of the target pixel X and the pixel value of the pixel A (FIG. 4), and generates a symbol based on the calculated difference value (prediction error value).
The encoding program 5 moves the position of the target pixel X in the scanning order to generate symbols for all the pixels. The generated symbols are output to the first encoding unit 510 and the symbol counting unit 520.

  In step 110 (S110), the encoding program 5 sets attention symbols (that is, symbols to be encoded) in order from the generated symbol group.

  In step 120 (S120), the first encoding unit 510 encodes the symbol of interest using the fixed code table 512 of FIG.

  In step 130 (S130), the intermediate code storage unit 530 buffers the symbol of interest (that is, the intermediate code) encoded by the first encoding unit 510 according to the buffering code format 514 of FIG. 5B.

  In step 140 (S140), the symbol counting unit 520 counts the symbols generated by the source coder 500 in parallel with the encoding process (S120) of the first encoding unit 510, and the appearance frequency distribution of each symbol value. Is generated.

  In step 150 (S150), the encoding program 5 determines whether or not all symbols have been processed. If all symbols have been processed, the process proceeds to S160, and there is an unprocessed symbol. , The process returns to S110.

  In step 160 (S160), the code table generation unit 540 generates a code table (variable code table) of the variable Huffman code according to the appearance frequency distribution of the symbol values generated by the symbol counting unit 520. The variable code table is output to the second encoding unit 550.

  In Step 170 (S170), the second encoding unit 550 encodes the attribute value of the intermediate code buffered by the intermediate code storage unit 530 using the variable code table generated by the code table generation unit 540, The encoded attribute value is used as an output code.

  As described above, the encoding apparatus 2 in the present embodiment can reduce the load required for symbol buffering by combining the fixed encoding process and the variable encoding process in series. Thereby, high-speed entropy encoding processing is realized.

[Modification 1]
Next, the modification 1 of the said embodiment is demonstrated.
In the above embodiment, the first encoding unit 510 and the second encoding unit 550 encode the run by applying the same maximum run length, but the present invention is not limited to this. That is, the encoding process by the first encoding unit 510 and the encoding process by the second encoding unit 550 may be slightly different.
For example, the maximum run length applied in the fixed encoding process (that is, the encoding process by the first encoding unit 510) is applied in the variable encoding process (that is, the encoding process by the second encoding unit 550). It may be different from the maximum run length.
In run length conversion, the maximum run length is a parameter. In the embodiment described above, the symbol counting unit 520 needs to count the appearance frequency of the run length 1 to the run length R in accordance with the maximum run length R in the variable processing because of applying the variable Huffman code. . Therefore, it is reasonable to make the run length of the maximum run length R by fixed encoding processing.
However, there is a case where the maximum run length may be increased by fixed encoding processing. For example, this is a case where image data whose prediction error value is 0 continues for a long time. In such a case, the data amount to be buffered can be reduced by setting the maximum run length in the fixed encoding process as long as possible.
At this time, the second encoding unit 550 needs to separate long runs into units of the maximum run length R, which is overhead. Accordingly, the run length may be increased as long as the overhead and the above-described effect are attracted. At this time, it is preferable that the maximum run length applied in the fixed encoding process is a multiple of the maximum run length applied in the variable encoding process because the variable process can be simplified.

FIG. 7 is a diagram for explaining processing when the maximum run length 256 is applied in the fixed encoding process and the maximum run length 32 of 32 is applied in the variable encoding process.
For example, when 300 runs illustrated in FIG. 7A are input to the first encoding unit 510, the first encoding unit 510 converts the 300 runs into 256 runs as illustrated in FIG. 7B. And 44 runs, each encoded as a symbol.
Then, the second encoding unit 550 divides 256 runs into 8 to generate 32 runs, divides 44 runs into 32 runs and 12 runs, and codes each divided run as a symbol. Turn into. In this case, the code table generation unit 540 generates a code table according to the appearance frequency distribution from run 1 to run 32, and the second code unit 550 uses this code table to identify the run as a variable Huffman code. Encode.

[Modification 2]
Next, a second modification of the above embodiment will be described.
In the above embodiment, the first encoding unit 510 and the second encoding unit 550 perform encoding processing in the same data unit (symbol unit), but the present invention is not limited to this. For example, the first encoding unit The processing unit (data unit) of the unit 510 may be larger than the processing unit (data unit) of the second encoding unit 550.
In the second modification, the first encoding unit 510 encodes a block composed of a plurality of pixels as a processing unit, and the second encoding unit 550 describes an embodiment in which one pixel is encoded as a processing unit. That is, in the above embodiment, since the identifier is assigned to each pixel in the fixed encoding process, 2 bits / pixel is required even for an all-white image, for example. In addition, since the identifier in units of pixels means switching of processing in units of pixels, there is a possibility that branch processing is frequently performed in the fixed encoding process and the variable encoding process, thereby reducing the processing speed.
Therefore, the first encoding unit 510 of the second modification example assigns an identifier to each block obtained by collecting a plurality of pixels. Since a recent CPU (central processing unit) can perform parallel processing by dividing a wide register, matching with a source coder to which such a technique can be applied is also good.

FIG. 8A illustrates a fixed code table 513 in the case of encoding in block units, and FIGS. 8B and 8C are applied when encoding an intermediate code generated in block units. It is a figure which illustrates the variable code table. In addition, in this modification, the form which makes 32 pixels 1 block is made into a specific example.
When all symbols corresponding to the pixels in the block are runs, the first coding unit 510 of this modification example sets the set of the identifier “00” and the run length (32 in this example) as the intermediate code of this block. And
In addition, when all symbols corresponding to the pixels in the block have prediction error values from −8 to 7, the first encoding unit 510 and the identifier “01” and the total prediction error value in the block (4-bit representation) )) Is the intermediate code of this block. The prediction error value (4 bits) included in the intermediate code may also include a prediction error value 0 (4-bit representation).
Further, the first encoding unit 510, when any of the symbols corresponding to the pixels in the block has a prediction error value smaller than −8 or a prediction error value larger than 7, the identifier “10” and this block The set of all of the prediction error values (8-bit representation) is the intermediate code of this block. Prediction error values (8-bit representation) from −8 to 7 can be included in the prediction error values (8-bit representation) included in the intermediate code.

Next, it is a code table used for variable encoding processing. As described above, the prediction error value 0 may be expressed in three types of run expression, 4-bit expression, and 8-bit expression. There is a possibility that the prediction error values (except 0) of −8 to 7 are expressed by two types of 4-bit expression and 8-bit expression.
Therefore, as illustrated in FIG. 8B, the code table generation unit 540 of the present modification includes an 8-bit input variable code table 552 including prediction error values from −8 to 7 as symbols, and a prediction error value of 0. Is generated as a symbol and a 4-bit input variable code table 554 is generated. In the 8-bit input variable code table 552, prediction error values (including 0) from −8 to 7 are expressed as 8-bit symbols, and the 4-bit input variable code table 554 includes the prediction error value 0 and 4 bits. It is expressed as a symbol.
The 8-bit input variable code table 552 and the 4-bit input variable code table 554 are look-up tables having 8-bit symbols or 4-bit symbols as entries, respectively. For example, a 256-entry lookup table may be prepared for 256 symbols as the 8-bit input variable code table 552, and a 16-entry lookup table may be prepared as the 4-bit input variable code table 554. As described above, in this modification, two lookup tables are required, but the influence is small because the number of entries in the 4-bit input variable code table 554 is small.

In the fixed encoding process, identifiers are assigned in units of blocks, and attribute values are in units of pixels. That is, an attribute value of 32 pixels is included in one block attribute value.
Therefore, even if the processing unit is a block, the data granularity is a pixel unit, and therefore, there is no mismatch between the processing unit and the variable encoding process.

[Other variations]
The encoding device 2 may generate a variable code table having a plurality of symbols as entries, and perform the variable encoding process using the variable code table, thereby speeding up the encoding process.
FIG. 9 is a diagram illustrating a 2-symbol variable code table 556 having 2 symbols as entries.
The second encoding unit 550 can increase the speed of the encoding process by encoding two symbols simultaneously using the two-symbol variable code table 556 illustrated in FIG. The 2-symbol variable code table 556 of this example has entries of prediction error values (symbol values 0 to 15) from −8 to 7. When image data is a target, symbols (prediction error values) are concentrated in the vicinity of 0, that is, the appearance frequency is high, and such speeding up is effective.

[Experimental result]
FIG. 10 is a graph of experimental results. The horizontal axis of the graph in this figure is the time (processing time) required for the encoding process, and the vertical axis is the code amount.
As shown in FIG. 10, when only the fixed Huffman code is applied, the processing time is short, but the code amount becomes large. On the other hand, when only the variable Huffman code is applied, the code amount can be reduced, but the processing time becomes longer.
When the fixed Huffman coding process is performed in units of blocks and the intermediate code is variable Huffman coded as in the second modification, a high compression rate can be realized in a relatively short processing time.

It is a figure which illustrates the function structure of the encoding program 9 to which a variable Huffman code is applied. It is a figure which illustrates the hardware constitutions of the encoding apparatus 2 with which the encoding method concerning this invention is applied centering on the control apparatus 20. FIG. It is a figure which illustrates the functional structure of the encoding program 5 which is performed by the control apparatus 20 (FIG. 2) and implement | achieves the encoding method concerning this invention. It is a figure explaining the source coding by the source coder 500 (FIG. 3). (A) illustrates the fixed Huffman code by the first encoding unit 510 (FIG. 3), and (B) illustrates the format of the intermediate code stored in the intermediate code storage unit 530 (FIG. 3). is there. It is a flowchart of the encoding process (S10) by the encoding program 5 (FIG. 3). It is a figure explaining the process in the case of applying the maximum run length 256 by fixed encoding processing, and applying the maximum run length 32 of 32 by variable encoding processing. (A) illustrates a fixed code table 513 in the case of encoding in block units, and (B) and (C) are variable code tables applied when encoding an intermediate code generated in block units. FIG. It is a figure which illustrates the 2 symbol variable code table 556 which uses 2 symbols as an entry. It is a graph of an experimental result.

Explanation of symbols

DESCRIPTION OF SYMBOLS 2 ... Encoding apparatus 5 ... Encoding program 500 ... Source coder 510 ... 1st code | symbol part 520 ... Symbol count part 530 ... Intermediate code memory | storage part 540 ... Code table production | generation Part 550 ... second code part

Claims (9)

First code generation means for encoding an input data file and generating an intermediate code using an entropy encoding method independent of unprocessed data;
And a second code generation unit configured to generate an output code of the input data file based on the intermediate code generated by the first code generation unit. The encoding apparatus according to claim 1, wherein the second code generation means generates the output code using an encoding method that depends on unprocessed data. The encoding apparatus according to claim 1, wherein the second code generation means generates the output code based on statistical processing for the entire data file. Statistical processing means for performing the statistical processing in parallel with at least part of the encoding processing by the first code generation means;
The encoding apparatus according to claim 3, wherein the second code generation means generates the output code based on a result of statistical processing by the statistical processing means. The encoding apparatus according to claim 1, wherein the first code generation unit performs an encoding process in a data unit larger than the second code generation unit. The first code generation means encodes the input data file using a pre-designed fixed code table,
The encoding apparatus according to claim 1, wherein the second code generation means generates the output code using a variable code table designed in accordance with each input data file. Storage means for storing the intermediate code generated by the first code generation means;
The encoding apparatus according to claim 1, wherein the second code generation unit generates the output code by encoding an intermediate code stored in the storage unit. Encode the input data file using an entropy encoding scheme that does not depend on raw data, generate an intermediate code,
An encoding method for generating an output code of an input data file based on the generated intermediate code. Encoding an input data file using an entropy encoding scheme independent of raw data to generate an intermediate code;
A program for causing a computer to execute an output code of an input data file based on the generated intermediate code.

Images