JP2000188692A - Data processing method - Google Patents

Abstract

PROBLEM TO BE SOLVED: To provide a data compression method that can compress compression objet data at high speed, from which a comparatively longer coincidence length is obtained. SOLUTION: When image data stream is coded and compressed by a plurality of lines whose line length is L, a specific symbol included in the image data stream and a plurality of symbols with the specific symbol placed at the head are used for coding object symbols, a symbol adjacent to an upper-stream side of the specific symbol with an offset of 1 and a plurality of symbols toward the upper-stream side and the lower-stream side of the stream with offsets of L+n to L-n around the symbol with an offset L, that is apart from the specific symbol by one line length, are used for comparison object symbols. The coding object symbols and the comparison object symbols are sequentially compared and detection of matching length is finished, when a maximum matching length is obtained which was specified in advance.

Description

DETAILED DESCRIPTION OF THE INVENTION

[0001]

[0001] 1. Field of the Invention [0002] The present invention relates to a data processing method, and more specifically, to LZ77 and LZ77.
The present invention relates to a data processing method for efficiently compressing image data by using a data compression technique based on a dictionary-based method represented by 78.

[0002]

2. Description of the Related Art Abraham Lempel and Jacob Ziv 197 the origin of the current dictionary-based data compression method.
Paper published in IEEE Transaction on Information Theory in 1995７AUniversal Algorithm for Sequential Data
Seen in Compression '. This is commonly called the slide dictionary method of Lempel-Ziv coding or the LZ77 method.

For example, Seiji Munakata: Data compression method of Ziv-Lempel, Information Processing, Vol. 26. No. 1 (1985)
It is introduced.

[0004] The LZ77 algorithm is a method in which coded data is divided from an arbitrary position in a past data sequence into a sequence of the maximum length that matches, and encoded as a duplicate of the past sequence.

More specifically, as shown in FIG. 2, a moving window for storing encoded input data and a look-ahead buffer for storing data to be coded from now on are provided. Is compared with all the subsequences of the data sequence, and a subsequence of the maximum length that matches in the moving window is obtained.

Then, in order to specify the maximum-length sub-sequence in the moving window, the "start position of the maximum-length sub-sequence", the "matching length", and the "next symbol that caused a mismatch" Is encoded.

Next, the encoded data sequence in the prefetch buffer is moved to the moving window, and a new data sequence corresponding to the encoded data sequence is input into the prefetch buffer.

Hereinafter, by repeating the same processing, the data is decomposed into sub-series and the encoding is executed.

[0009] Many improvements have been proposed for such basic data compression techniques.

For example, an LZSS code which employs a method of providing a flag for identifying whether it is an encoded code or raw data and encoding the raw data when the encoded code becomes longer than the raw data. Method (TCBel
l, “Better OPM / L TextCompression”, IEEE Transaction
Commun., Vol. COM-34, No. 12, Dec. (1986)).

Further, as another document, M. Nelson:
Data Compression Handbook, 2nd revised edition, Toppan (1996).
ISBN4-8101-8605-9.

In recent years, OA systems (scanners, printers, digital copiers, etc.) have become widespread and aiming for higher speed and higher resolution.

In these apparatuses, it is necessary to process a large amount of image data at high speed, and it is essential to reduce the amount of data to be processed by applying high-speed and high-compression data compression. .

As a conventional technique for such data compression, there are standardized methods such as MMR and JBIG.
In the MMR, the compression ratio tends to be deteriorated in a fine image.

Also, JBI which is close to the best in terms of compression ratio
Since G is basically processing in units of pixels, there is a limit in speeding up, and it cannot be adopted in a high-speed system.

However, the dictionary-based compression method described above is basically a processing in units of bytes, so that it is possible to achieve a much higher speed than JBIG, and that the compression rate does not deteriorate as much as MMR even for a fine image. It has features and is suitable for high-speed and high-resolution OA systems.

[0017]

However, in a conventional LZ77-based data compression apparatus, when encoding, in order to obtain a partial data string having the same maximum length in a moving window, a data string to be encoded from now on is required. A data string comparison must be performed between the and all positions in the moving window.

That is, as shown in FIG. 2, a data sequence to be encoded is a data sequence starting from the position of offset 1 in the moving window, a data sequence starting from the position of offset 2,. Is compared with the data sequence starting from the position of (size) to find the offset that gives the maximum match length.

The method of obtaining the maximum match length as described above has a disadvantage that the processing speed is reduced when a long match is obtained with each offset. For example, encoding the white part of the image gives the longest match (eg 256) in comparison with all offsets, so counting one data comparison as one would result in 256 comparisons for each offset Therefore, there is a problem that the comparison time of the data sequence is dramatically increased.

SUMMARY OF THE INVENTION An object of the present invention has been made in view of the above circumstances, and enables high-speed compression of compression target data in which identical data is continuous, that is, compression target data having a relatively long matching length. Another object of the present invention is to provide a simple data compression method.

[0021]

In order to solve the above problems and achieve the object, a data compression method according to the present invention is as follows.

The present invention provides a data processing method for encoding an image data stream corresponding to a plurality of lines having a line length L of the image obtained by scanning the image and compressing the image data stream. A specific symbol included and a plurality of symbols starting with the specific symbol are set as encoding target symbols, and a symbol having an offset of 1 adjacent to the upstream side of the specific symbol and a distance of one line length from the specific symbol are used. And a plurality of symbols at offsets L + n to L-n on the upstream side and the downstream side of the stream centered on the symbol at the offset L, are set as comparison target symbols, and the encoding target symbol and the comparison target symbol are sequentially compared. When a match length is detected, the point in time when a predefined maximum match length is obtained Exit matching length detection, encoding is performed by the detected match length based.

The present invention provides a data processing method for encoding an image data stream corresponding to a plurality of lines having a line length L of the image obtained by scanning the image and compressing the image data stream. A specific symbol included and a plurality of symbols starting with the specific symbol are set as encoding target symbols, and a symbol having an offset of 1 adjacent to the upstream side of the specific symbol and a distance of one line length from the specific symbol are used. And a plurality of symbols at offsets L + n to L-n on the upstream side and the downstream side of the stream centered on the symbol at the offset L, are set as comparison target symbols, and the encoding target symbol and the comparison target symbol are sequentially compared. When encoding, the specific sequence included in the symbol to be encoded is included. As the object of comparison of Bol, the offset 1, L, L + 1, and preferentially selects the L-1 symbols.

According to the present invention, there is provided a data processing method for encoding an image data stream corresponding to a plurality of lines having a line length L of the image obtained by scanning the image and compressing the image data stream. A specific symbol included and a plurality of symbols starting with the specific symbol are set as encoding target symbols, and a symbol having an offset of 1 adjacent to the upstream side of the specific symbol and a distance of one line length from the specific symbol are used. And a plurality of symbols at offsets L + n to L-n on the upstream side and the downstream side of the stream centered on the symbol at the offset L, are set as comparison target symbols, and the encoding target symbol and the comparison target symbol are sequentially compared. When encoding, the specific sequence included in the symbol to be encoded is included. The comparison priority Bol in descending order of priority, the offset L, 1, L-1, L
+1, L−2, L + 2,..., L−n, and L + n, and when the encoding target symbol and the comparison target symbol are sequentially compared to detect a matching length, a predetermined maximum matching length is determined. Is detected at the time point when is obtained, and encoding is performed based on the detected match length.

[0025]

Embodiments of the present invention will be described below with reference to the drawings.

First, the point at which a comparison point is selected from two-dimensionally close positions will be described.

As described in the conventional example, when realizing LZ77-based compression by software, in the simplest implementation, a data sequence starting from a code or a position and a data sequence starting from all positions in a moving window are obtained. And the longest matching position is detected. In this method, when the moving window is large, the processing speed is significantly reduced.

Therefore, in the first invention, based on LZ77, a data sequence to be encoded is not compared with a data sequence starting from all positions in the moving window. The processing speed is improved by making only the data sequence starting with "." For example, it is realized at about 16 or 32 as comparison target positions.

However, simply reducing the number of comparison target positions is considered to reduce the possibility of coincidence and reduce the compression ratio. In the first invention, a position to be compared is selected by focusing on the periodicity of the image data.
That is, instead of comparing all the partial columns in the moving window, attention is paid to the image data periodicity, and only data positions that are highly likely to match are compared.

The principle of the first invention will be described below with reference to FIG. Here, the unit of data to be compressed is a byte unit. In consideration of the two-dimensional locality of the image data, the highest similarity to a certain byte is the upper, lower, left, and right positions. Considering a typical raster scan from upper left to lower right as the input order of image data, for a certain byte, the right and lower adjacent bytes are to be input from now, so they have not yet entered the moving window . Therefore, the right and lower adjacent bytes cannot be compared.
The left adjacent byte is the most recent data that was input immediately before in the input order and is located at offset 1 in the moving window. The upper adjacent byte is located at the offset L in the moving window, where the line length (width) of the input image data is L in bytes. However, the condition is that the size of the moving window is L or more. In the conventional LZ codec, even if the input is image data, it ignores the periodicity and searches for a matching point only in the left direction. Here, in addition to the left direction, a position adjacent in the upward direction and its periphery are selected as comparison points. In FIG. 1, a diamond indicates a first byte of data to be encoded from now on, an equal indicates a data sequence to be encoded from now on, a black square indicates a byte position as a comparison point in the moving window, and a margin is shown. Indicates a byte position that is not a comparison point in the moving window. That is, here, 16 offset positions (1, L-7, L-6, L-5, L-4, L-3, L-
2, L-1, L, L + 1, L + 2, L + 3, L + 4, L
+5, L + 6, L + 7). L is the line length of the image (the size of the main scanning method) and is set in advance from the outside.

Reducing the number of comparison targets is also advantageous in that the offset code can be shortened. For example, if the size of the moving window is 2 kB, the conventional example requires 2 k kinds of offset codes, but the example of FIG. 1 requires only 16 kinds of offsets. In this case, the offset code length is 11 bits, whereas in the present invention, it is as short as 4 bits.

Further, in the present invention, the order of obtaining the match length at each offset and the condition for terminating the longest match search are combined to speed up the compression processing. The method of simply finding the longest match for the 16 comparison points of the present invention is as follows.

A module for simply obtaining the longest match for the 16 comparison points. {The match length between the data sequence from the encoding position and the data sequence from offset 1 is obtained, and the result is len1. The data from the encoding position The matching length of the data string from the column and the offset L-7 is determined, and the result is len (L-7). The matching length of the data string from the encoding position and the data string from the offset L-6 is determined. Let len (L-6) ... Find the matching length of the data string from the encoding position and the data string from offset L + 6, and let the result be len (L + 6). The data string from the encoding position and the data from offset L + 7 The match length of the column is determined, and the result is len (L + 7). Len1, len (L-7), (L-6), len
Returns the maximum value of (L + 7) and the offset at that time. When determining the match length at each offset, the upper limit is determined by the configuration of the match length code, and even if a long match is obtained, the comparison process is terminated at the upper limit. . For example, when the maximum length of the match length code is 256, when the number of matches reaches 256, the subsequent comparison is stopped and 256 is set as the match length.

The conventional method of obtaining the maximum matching length has a disadvantage that the processing speed is reduced when a long matching is obtained for each offset. For example, encoding a white portion of an image will yield the longest match (eg, 256) in comparison with all offsets, so counting one data comparison as one would result in 256 times for each offset, for a total of 2096 times A comparison will be made.

In the present invention, attention is paid to the fact that 256 is the maximum matching length, and when the matching length of 256 is obtained, the comparison of the remaining offsets is cut off, thereby increasing the speed. However, since the Huffman code is usually used for both the offset code and the match length code, a short offset code is assigned to the offset at which a long match length is likely to be obtained. For example, when an offset is selected as in the present invention, it is considered that the closer to the position to be coded from now on has higher similarity, so the shortest code is assigned to L, and L-
1, L + 1, L-1, L + 1, L-2, L + 2 ..., L-
It is preferable to assign short codes in the order of 7, L + 7. At this time, if a truncation is simply introduced with the maximum matching length as follows, a non-shortest offset may be selected, which is not optimal.

Introduce truncation into the module that simply finds the longest match for the 16 comparison points of the present invention. 求 め Find the match length between the data string from the encoding position and the data string from offset 1, and call the result len1 If len1 = 256, offset = 1, match length = 25
The module length is set to 6, and the matching length of the data sequence from the encoding position and the data sequence from the offset L-7 is determined, and the result is len (L-7). If len (L-7) = 256, the offset = L− 7,
Module length is set as the match length = 256. The match length between the data string from the encoding position and the data string from the offset L-6 is obtained, and the result is len (L-6). If len (L-6) = 256, the offset is obtained. = L-6,
The module ends with a match length = 256. The match length between the data string from the encoding position and the data string from the offset L + 6 is determined, and the result is len (L + 6). If len (L + 6) = 256, the offset = L + 6,
Module length is set as the match length = 256. The match length between the data string from the encoding position and the data string from the offset L + 7 is determined, and the result is len (L + 7). If len (L + 7) = 256, the offset = L + 7,
End of module with matching length = 256 len1, len (L-7), (L-6) ..., len
Return the maximum value of (L + 6) and len (L + 7) and the offset at that time. When searching in this order, the white part of the image has an offset of 1
Is selected, but since the shortest code is assigned to the offset L, it is not the optimum code. In order to improve this point, the search order may be set in the order of the shortest offset code (not the longest order) as follows.

Improving the search position by introducing truncation (method of the present invention) 求 め Find the coincidence length of the data sequence from the encoding position and the data sequence from the offset L and set the result to lenL If lenL = 256, the offset = L, match length = 25
The module length is determined as 6, and the matching length between the data sequence from the encoding position and the data sequence from the offset L is determined. The result is lenL. If len = 1, the offset = 1 and the matching length = 256. The matching length between the data sequence and the data sequence from the offset L-1 is determined, and the result is len (L-1). If len (L-1) = 256, the offset = L-1,
Module length is set as the match length = 256. The match length between the data string from the encoding position and the data string from the offset L + 1 is obtained, and the result is len (L + 1). If len (L + 1) = 256, the offset = L + 1,
When the matching length = 256, the module ends ... The matching length between the data string from the encoding position and the data string from the offset L-7 is determined, and the result is len (L-7). If len (L-7) = 256, Offset = L-7,
Module length is set as the match length = 256. The match length between the data string from the encoding position and the data string from the offset L + 7 is determined, and the result is len (L + 7). If len (L + 7) = 256, the offset = L + 7,
End of module with matching length = 256 len1, len (L-7), (L-6) ..., len
Returns the maximum value of (L + 6) and len (L + 7) and the offset at that time. In this case, even if 256 matches occur and the search for a matching length thereafter is interrupted, the offset with the shortest offset code is always selected. Therefore, it is possible to improve the encoding efficiency and increase the processing speed at the same time. For example, for the white part of the image, the initial method required a total of 4096 comparisons, but the method of the present invention requires 256 comparisons, and the same sign is generated.

The search termination condition of the present invention may be set to be smaller than the maximum value of the matching length code in order to realize high speed even if the coding efficiency is slightly lowered. For example, when the maximum value of the match length code is 256, if a match exceeding 128 is obtained, the subsequent search is not performed, so that although the coding efficiency is slightly reduced, the speed can be increased.

Next, the second invention will be described.

In the module for obtaining the longest matching position, the data sequence from the coding position is compared with the data sequence starting from each offset position to determine the matching length. In the present invention, a total of 16 offsets are calculated. Comparison is made. When the encoding unit is byte-wise, if it is simply implemented, conventionally, as shown below, 1
You will compare byte by byte. This may be the case where the compression processing is performed by an 8-bit CPU, but considering the recent implementation in a 32-bit CPU, the speed can be increased as follows. Assuming that the natural data length of the CPU is 32 bits, the second invention implements the following as described below. int search_maechlen (BYTE * offset, BYTE * cp) {count = 0 (If the difference between offset and cp is not a multiple of 4 bytes, compare in 1-byte units as in the conventional example, otherwise perform the following processing (Offset and cp compare 1 to 3 bytes until they match the 4-byte boundary) while (* (int) offset == * (int) cp) compare in count + = 4 // 4 byte units (4 bytes of last mismatch) 1 to 3 bytes match, or if they match, add count for that match) return (count)} Unless the offset is a multiple of 4 bytes, comparison in 4-byte units is not possible, so 1 byte as in the past Compare by unit. If the offset is a multiple of 4 bytes, the part that matches the 4-byte boundary can be compared in 4-byte units. First, 1 to 3 bytes are compared in 1-byte units to determine whether they match up to the 4-byte boundary. After matching with the 4-byte boundary, high-speed comparison is performed in 4-byte units. Continue until a mismatch occurs or the maximum match length code is reached. Even when a mismatch occurs, there is a possibility that 1 to 3 bytes match within the last 4-byte boundary, so comparison is made in byte units.

In a general 32-bit CPU, 1-byte comparison and 4-byte comparison are processed in the same cycle. Therefore, by performing comparison as in the present invention, up to four times the speed can be realized.

Next, the third invention will be described.

In the decompression process, a code is decoded to obtain a match offset and a match length, data is copied from the match offset in the data buffer by the match length, and added to the data output as new decoded data. Become. At this time, if simply implemented, conventionally, memory copying is performed one byte at a time as shown below. void matchl_copy (BYTE * offset, BYTE * cp, int length) {memcpy_in_BYTE (cp, offset, length)} Similar to the second invention, the C of 32 bits or the like is used recently.
Considering the implementation on the PU, the speed can be increased as follows. CP
Assuming that the natural data length of U is 32 bits, the third embodiment implements the following as described below. void matchl_copy (BYTE * offset, BYTE * cp, int length) {(If the difference between offset and cp is not a multiple of 4 bytes, copy in 1-byte units as in the conventional example. Otherwise, perform the following processing. (Offset and cp are copied 1 to 3 bytes until they match the 4-byte boundary) memcpy_in_4BYTE (cp, offset, length) / Memory copy in units of 4 bytes (if there is a remainder, copy 1 to 3 bytes)} If the offset is not a multiple of 4 bytes, memory copy cannot be performed in units of 4 bytes. Therefore, memory copy is performed in units of 1 byte as in the conventional example. If the offset is a multiple of 4 bytes, the portion that matches the 4-byte boundary can be memory-copied in 4-byte units. After matching the 4-byte boundary, high-speed memory copying is performed in 4-byte units until the matching length is reached. Finally, if there is not a 4-byte boundary and there are too many, memory copy is performed in 1-byte units.

In a general 32-bit CPU, 1-byte memory copy and 4-byte memory copy are processed in the same cycle. Therefore, by performing memory copy as in the third aspect of the invention, up to four times the speed can be realized.

Next, a list of points of the present invention will be summarized.

When expressing the compression / expansion of the Lempel-Ziv method (moving window method) by software <Compression> If a long match is obtained with the offset checked earlier, other offsets are not checked. (Conventional) for (from offset1 to offset N) {len 1 = search_matchlen (offset 1, current_pointer) len 2 = search_matchlen (offset 2, current_pointer) len 3 = search_matchlen (offset 3, current_pointer)…} maxlen = max (len 1 , len 2 ...) (the present invention) for (from offset 1 to offset N) {if (len 1 = search_matchlen (offset 1, current_pointer)> = thresh_len) break if (len 2 = search_matchlen (offset 2, current_pointer) > = Thresh_len) break if (len 3 = search_matchlen (offset 3, current_pointer)> = thresh_len) break…} maxlen = max (len 1, len 2 ...) The search_matchlen function is used as the native word length (32-bit 4 bytes). <Expansion> When forming an original image from a matching code, if the word boundaries match, the memcpy operation is executed by a 4-byte copy instruction.

As described above, according to the present invention, when the LZ77-based compression apparatus is realized by software, the offset which gives the longest match is efficiently searched by paying attention to the periodicity of the image data, and the longest match is given. When the offset is found, the search for the matching length ends. Thereby, an extra matching length search process can be omitted, and the compression processing speed can be improved. When checking the matching length at each offset, for example, the unit of compression (1 byte)
When the native word length (for example, 4 bytes in the case of a 32-bit processor) of the processor that implements the compression software differs, the processing efficiency can be improved by comparing the match with the native word length of the processor. Further, when the decompression processing is implemented by software, the processing efficiency can be improved by performing memory copy with the native word length of the processor in decoding the coincidence code.

[0048]

According to the present invention, it is possible to provide a data compression method capable of rapidly compressing compression target data in which the same data continues, that is, compression target data having a relatively long matching length.

(1) When a conventional LZ77-based data compression apparatus is implemented by software, a data sequence to be encoded is compared with a data sequence to be compared and all offset positions to be compared. Unnecessary comparisons were made in areas consisting entirely of white, such as the periphery. According to the present invention, attention is paid to the periodicity of image data, comparison is performed from offset positions that are two-dimensionally close to each other, and when an offset position that reaches the maximum matching length defined by the matching code is obtained, comparison with other offset positions is performed. By canceling, the compression processing time can be reduced.

Even if the maximum match length specified by the match code is not reached, a threshold value of the match length is set in advance, and if a match length longer than that is obtained, the comparison with the subsequent offset is canceled. A similar effect is obtained. In this case, an optimal compression ratio may not be obtained. However, by adjusting the threshold value, a decrease in the compression ratio can be suppressed.

(2) When a conventional LZ77-based data compression apparatus is implemented by software, it is necessary to compare a data string to be encoded from now with a data string from an offset position to be compared, and check a matching length. . Conventionally, when the processing unit of compression is 1 byte, both data strings are compared in 1-byte units to determine the matching length. According to the present invention, the native word length (for example, 3 bytes) of a processor that implements compression software is used instead of the compression unit (for example, 1 byte).
If the 2-byte processor is large (4 bytes), processing efficiency is improved by comparing data strings with the native word length of the processor.

(3) When a conventional LZ77-based data decompression device is implemented by software, it is necessary to copy the original data of the matching length from the offset position indicated by the matching code in decoding the matching code. Has performed memory copy in a compression processing unit (for example, 1 byte). In the present invention, when the native word length of a processor implementing compression software (for example, 4 bytes in the case of a 32-bit processor) is larger than the unit of compression (for example, 1 byte), memory copy is performed using the native word length of the processor. This improves the processing efficiency.

[Brief description of the drawings]

FIG. 1 is a diagram for explaining an outline of a data compression method according to the present invention.

FIG. 2 is a diagram for explaining an outline of a conventional data compression method.

Claims (7)

[Claims] 1. A data processing method for encoding an image data stream corresponding to a plurality of lines having a line length L of the image obtained by scanning the image and compressing the image data stream, wherein the image data stream is included in the image data stream. Certain symbols,
And a plurality of symbols starting from the specific symbol as encoding target symbols, a symbol having an offset 1 adjacent to the upstream side of the specific symbol and a symbol having an offset L one line length away from the specific symbol. When a plurality of symbols at offsets L + n to L-n on the upstream and downstream sides of the centered stream are set as comparison target symbols, and the encoding target symbols and the comparison target symbols are sequentially compared to detect a coincidence length. A data processing method for terminating detection of a match length when a predetermined maximum match length is obtained, and performing encoding based on the detected match length. 2. A data processing method for encoding an image data stream corresponding to a plurality of lines having a line length L of the image obtained by scanning the image and compressing the image data stream, wherein the image data stream is included in the image data stream. Certain symbols,
And a plurality of symbols starting from the specific symbol as encoding target symbols, a symbol having an offset 1 adjacent to the upstream side of the specific symbol and a symbol having an offset L one line length away from the specific symbol. When a plurality of symbols at an offset L + n to an offset L−n on the upstream side and the downstream side of the center stream are set as comparison target symbols, and the encoding target symbol and the comparison target symbol are sequentially compared and encoded, The offsets 1, L, L + 1, and L-1 may be compared with the specific symbol included in the encoding target symbol.
A data processing method characterized by preferentially selecting a symbol. 3. A data processing method for encoding an image data stream corresponding to a plurality of lines having a line length L of the image obtained by scanning the image and compressing the image data stream, wherein the image data stream is included in the image data stream. Certain symbols,
And a plurality of symbols starting from the specific symbol as encoding target symbols, a symbol having an offset 1 adjacent to the upstream side of the specific symbol and a symbol having an offset L one line length away from the specific symbol. When a plurality of symbols at an offset L + n to an offset L−n on the upstream side and the downstream side of the center stream are set as comparison target symbols, and the encoding target symbol and the comparison target symbol are sequentially compared and encoded, The offsets L, 1, L-1, L + 1, L-2, L + 2,..., L- are set in order of comparison priority of the specific symbols included in the encoding target symbols, in descending order of priority.
n and L + n symbols, and sequentially comparing the encoding target symbol with the comparison target symbol to detect a match length, ends the match length detection when a predetermined maximum match length is obtained. And performing encoding based on the detected matching length. 4. The data processing apparatus according to claim 1, wherein the data processing unit compresses the data in units of n bits when the processing unit is an n-bit processing unit. Data processing method. 5. When one symbol is 8-bit data, and the arithmetic processing unit responsible for compression processing is an n-bit arithmetic processing unit, the data is compressed in units of (n / 8 × symbol). The data processing method according to claim 1, wherein the data is processed. 6. An arithmetic processing unit for copying data when a part of data included in an encoded data stream is copied in order to decompress an encoded data stream. 6. The data processing method according to claim 1, wherein the data is copied in units of n bits. 7. When copying a part of data included in an encoded data stream in order to decompress the encoded encoded data stream, one symbol is 8-bit data and the data is copied. 5. The data processing apparatus according to claim 1, wherein when the arithmetic processing unit is an n-bit arithmetic processing unit, data is copied in units of (n / 8.times.symbol). Item 6. The data processing method according to Item 5.