JP3193140B2 - Image and code data compression - Google Patents

Description

DETAILED DESCRIPTION OF THE INVENTION

[0001]

BACKGROUND OF THE INVENTION 1. Field of the Invention The present invention relates to an image and code data compression method for selectively compressing binary image data and code data such as characters.

[0002]

2. Description of the Related Art When data is transmitted or stored, data compression is often performed. As a coding method for performing data compression, an arithmetic code or a Ziv-Lempe (Ziv-Lempe) which is one of predictive coding methods is used.
The universal code of l) is known.

[0003] Arithmetic codes are well-suited for data compression of binary image data because they are suitable for data of Markov information sources and can be adapted according to the characteristics of the data. On the other hand, the universal code has a high compression effect on data in which long data of the same symbol pattern repeatedly appears, and is therefore often used for data compression of code data such as character codes.

There is a case where it is desired to arbitrarily select either image data or code data and compress the data.

[0005] In this case, conventionally, two types of coding schemes, such as the above-mentioned arithmetic code and universal code, have been selectively used. For this reason, two types of encoding means must be provided in the device, and the device is complicated.

On the other hand, for example, as disclosed in Japanese Patent Application Laid-Open No. 3-70268, there has been proposed a technique in which image data and character code data are compressed by one encoding means.

[0007] In this proposal, character code data is subjected to data compression by a universal code in byte units.
The image data is MH (Modified Huf).
fman), MR (Modified Relativ)
e Element Address Designna
te) or MMR (Modified MR) method, and then assigns a fixed-length code to the run-length code and mode information obtained by the encoding.
The fixed-length code is data-compressed by a universal code. By this processing, the periodicity of each data of the character code and the fixed length code is absorbed, and the compression effect becomes relatively high.

However, in the above proposal, MH, MR
Alternatively, since another encoding means called MMR is required, the apparatus is complicated as described above.

In the case of the universal code, when long data of the same symbol sequence repeatedly appears in the data to be processed, the compression effect is improved, but the data length of the same symbol sequence is short or the number of appearances is small. If
The compression effect was reduced.

[0010]

As described above, conventionally,
If a high compression effect is to be obtained for both the image data and the code data, a plurality of coding means are required, resulting in a problem that the apparatus becomes complicated.

An object of the present invention is to solve the above problems and provide a method for compressing both image data and code data with a simple apparatus configuration, and a high-compression image and code data compression method. Aim.

[0012]

According to the present invention, for this purpose, a predetermined number of input data are arranged in parallel, a symbol appearance probability of the bit of interest is predicted from the state of each bit surrounding the bit of interest, and the predicted value and A known arithmetic encoding unit for compressing data by encoding the correspondence with the actual symbol is 1
When the binary image data is compressed, the number of pixels in one line of the image is set as the fixed number, and the data is compressed by the arithmetic coding means. Sets the number of bits per code data as the fixed number, and compresses the data by the arithmetic coding means.

[0013]

Since only one encoding means is required, the structure of the apparatus is simplified. Further, since the image data is compressed as before by the arithmetic coding means, a high compression effect can be obtained. Further, when predicting the symbol appearance probability, the code data is arranged one by one in parallel with the code data, so that the correlation between the target bit and each surrounding bit becomes strong. As a result, the prediction accuracy is improved, and a high compression effect can be obtained even for code data.

[0014]

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

FIG. 1 is a block diagram showing an encoding apparatus according to one embodiment of the present invention. In the figure,
The P / S converter 1 converts character code data of a parallel signal into data of a serial signal. The input switching means 2 switches between inputting image data and inputting character code data. The reference symbol acquiring means 3 is for arranging the input data in a fixed number at a time and determining the state of each bit around the bit of interest. The reference symbol acquiring means 3 includes a one-line bit number switching means 3a for switching the above-mentioned fixed number for arranging the input data in two stages in two stages.

The probability table 4 is a data table that stores various probability values indicating the appearance probabilities of symbols.
The predicted value / probability data selection table 5 is a data table that stores a predicted value of a symbol and a pointer pointing to one probability value in the probability table 4 corresponding to each of the determined bit states. is there. The prediction determining means 6 determines whether the predicted value of the symbol matches the actual value, that is, whether or not the prediction was successful. In addition, this determination result indicates, in other words, whether the data symbol is an MPS (dominant symbol) or an LPS (inferior symbol). The arithmetic code generation means 7 generates an arithmetic code based on the determination result and the read probability value.

The table rewriting means 8 rewrites the pointer in the predicted value / probability data selection table 5 according to the hit state of symbol prediction. The identification code insertion means 9 inserts an identification code indicating which data is prior to each of the encoded data of the image and the character.

FIG. 2 shows a block diagram of a decoding apparatus for restoring the encoded data generated by the above-mentioned encoding apparatus to the original data. In the figure, an identification code detection / removal means 10 detects an identification code of input encoded data and removes the identification code. The arithmetic code decoding means 11 determines whether the symbol of the data is MPS or not by using the encoded data and the probability value.
S is determined.

The reference symbol acquisition means 3, probability table 4, prediction value / probability data selection table 5, prediction determination means 6, and arithmetic code generation means 7 have the same functions as those in FIG. Note that, in this case, the prediction determination means 6
The symbol of the data is reproduced based on the determination result of the LPS and the symbol prediction value.

The output switching means 12 switches the signal path between when the image data is reproduced and when the character code data is reproduced. The S / P converter 13 converts serial signal data into parallel signals.

Next, the operation of the encoding apparatus according to the present embodiment having the above configuration will be described.

Image data or character code data is arbitrarily input to the encoding device together with the data type signal. The data type signal indicates whether the input data is image data or character code data. The image data is binary data obtained by executing a main scan and a sub scan at a fixed resolution on a document image, and is sequentially input line by line. The character code data is character string information such as alphanumeric characters, and is sequentially input in the form of an ASCII code.

As shown in FIG. 3, when a data type signal is input, the coding apparatus determines the type of data input by the signal (process 101). Now, for example, if the input data is image data (processing 1
01, “image”), the image data side is selected by the input switching means 2 (processing 102), and an identification code indicating an image is output from the identification code inserting means 9 (processing 103).
Then, the one-line bit number switching means 3a sets the number of bits of one line to the number of pixels in the main scanning direction of the original image (process 104).

On the other hand, if the input data is character code data ("character" in processing 101), the input switching means 2 selects the P / S conversion means 1 (processing 105). An identification code indicating a character is output (process 106). Then, the one-line bit number switching means 3a sets the number of bits of one line to 8 bits, which is one unit of the ASCII code (process 107).

Then, a predetermined encoding process is started. That is, when image data is input, the image data is input to the reference symbol acquisition unit 3 as it is. When character code data is input, P / S
The signal is converted into a serial signal by the conversion means 1 and input to the reference symbol acquisition means 3.

In the reference symbol acquiring means 3, as shown in FIG. 4, input data is sequentially arranged in parallel by the set number of bits per line. Then, paying attention to each input bit, according to a preset template T,
The 10-bit data at a fixed position with respect to the target bit X is extracted. That is, the bits to be extracted are 2-bit data D1 and D2 on the same line, 5-bit data D2 to D6 on the previous line, and 3-bit data D7 to D9 on the line immediately before the previous line.

Since the extracted data D0 to D9 are 10 bits, as shown in FIG. 5A, the states of the symbol patterns are arranged in 1024 patterns. The reference symbol acquisition means 3 determines the symbol pattern of the extracted data D0 to D9 as states 0 to 1023.

The predicted value / probability data selection table 5 includes:
As shown in FIG. 13B, the predicted value of the MPS and the pointer value are stored corresponding to each of the states 0 to 1023 of the symbol pattern. The predicted value / probability data selection table 5 outputs a predicted value and a pointer value corresponding to the one determined state. The probability table 4 stores various probability values to which serial numbers are assigned, as shown in FIG. The probability table 4 outputs the probability value of the number indicated by the pointer value.

The prediction judging means 6 judges whether or not the prediction is correct by comparing the predicted value of the MPS with the symbol of the input data. In other words, this determination result indicates that the symbol of the input data is MPS or LP
S. The arithmetic code generating means 7 generates an arithmetic code by a known operation based on the determination result and the probability value, and outputs the generated arithmetic code to the outside (the above processing 10).
8).

In parallel with the above encoding operation, the table rewriting means 8 counts the number of times the symbol prediction was successful and the number of times the symbol prediction was missed (step 109). Then, it is determined whether or not the counted number has reached a predetermined number (step 110). If the specified number of times has not been reached (N in process 110), it is then determined whether or not data switching has been instructed by the data type signal (process 111). If data switching is not instructed (N in process 111), it is determined whether the data is completed (process 112). If not (N in process 112), the operation is continued (to process 108).

Here, it is assumed that image data has been input. In this case, in the reference symbol acquiring means 3, the image data is arranged line by line as shown in FIG. 4, and data DO to D9 around the target bit X is extracted by the template T. In the case of image data, it is well known that the correlation between these data DO to D9 and the target bit X is strong. In the above encoding process,
The symbol of the bit of interest X is converted to data DO to D having a strong correlation.
9, the prediction accuracy is high and the data compression ratio is high.

Next, it is assumed that character code data has been input. In this case, in the reference symbol acquisition means 3,
The character code data is arranged in order with 8 bits as one line, and each data DO to D9 around the target bit X is extracted in the same manner as described above.

Assume that the input character code data is character string data "ASCII ...". The ASCII code is 8 bits per character,
When they are arranged in the reference symbol acquiring means 3, they become as shown in FIG. That is, “A” is “0100000”
1 "," S "is" 01010011 "," C "is" 01 "
“0010011” and “I” are data “01001001”, and the upper 3 to 4 bits of each character are often identical. Therefore, the correlation between the data DO to D9 extracted from the template T and the target bit X is strong.

Thus, when encoding the character code data, the prediction accuracy is increased and the data compression ratio is increased, as in the case of the image data.

On the other hand, if the number of hits or misses of the symbol prediction reaches the specified number (Y in step 110), the pointer value stored in the predicted value / probability data selection table 5 is rewritten. That is, when the number of hits reaches the specified number, the pointer value is rewritten so as to indicate a higher probability value in the probability table of the probability table 4. On the other hand, when the number of departures reaches the specified number, the pointer value is rewritten so as to indicate a lower probability value (process 113). As a result, adaptation according to the characteristics of the data currently being encoded is performed, and the symbol prediction accuracy is further improved.

Next, it is assumed that data switching is instructed by the data type signal. In this case (Y in step 111), the data type is determined, and the above processing is executed similarly (to step 10). As a result, as shown in FIG. 7, the encoded data of each of the image and the character is sequentially output together with the identification code indicating the data type. Then, when there is no more input data (Y in process 112), the above-described encoding process ends.

Next, the operation of the decoding apparatus of this embodiment will be described.

The data signal encoded by the encoding device is input to the decoding device. As shown in FIG. 8, when a data signal is input to the decoding device, the identification code detecting / removing means 10 reads the identification code and determines the data type (process 201).

If the data signal is image data ("image" in step 202), the bit number of one line is set by the one-line bit number switching means 3a to the number of pixels in the main scanning direction of the original image ( Process 203). If it is character code data ("character" in step 202), the number of bits per line is set to 8 bits by the one-line bit number switching means 3a (step 204). Then, in the case of image data, the signal line for image output is selected by the output switching means 12, and in the case of character data, the S / P conversion means 13 is selected (process 205).

Then, a predetermined decoding process is executed. That is, the arithmetic code decoding means 11 decodes the information into whether the original data symbol is MPS or LPS based on the input coded data and the probability value output from the probability table 4. The prediction determination unit 6 restores the original data symbol based on the information and the predicted value output from the predicted value / probability data selection table 5.
In the case of image data, the restored data is output as it is. In the case of character code data, the data is converted into a parallel signal by the S / P converter 13 and output. In addition,
The reference symbol acquisition means 3, the predicted value / probability data selection table 5, the probability table 4, and the arithmetic code generation means 7 operate in the same manner as in the case of the above-described encoding apparatus.
Process 206).

In parallel with the above decoding operation, as in the case of FIG. 3, the symbol prediction result is determined, and the pointer value of the prediction value / probability data selection table 5 is rewritten according to the determination result (processing 109, process 110 and process 1
13).

During the above operation, the identification code of the input data is monitored (process 207). If the identification code is not detected (N in process 207), it is determined whether or not the input data is completed (process 208). If not (N in process 208), the decoding operation is continued (process 206).
What). Then, when there is no more input data (Y in process 208), the above-described decoding process ends.

As described above, the encoding apparatus according to the present embodiment
One encoding means for arithmetic codes is provided, and the one encoding means compresses both image data and character code data.

As a result, it is not necessary to provide two types of encoding means as in the prior art, so that the apparatus configuration is simplified.
In addition, since the decoding device only needs to provide one decoding means for the corresponding arithmetic code, the device configuration is similarly simplified.

Since the arithmetic code is a known coding method suitable for image data, a high compression effect can be obtained.
When the character code data is subjected to data compression, the input data is arranged in parallel in the reference symbol acquisition means 3 in units of 8 bits which is the code length of one character, and each bit data around the target bit is extracted. I have to.

In the case of character code data, even if the character is different, the bit at the same position is likely to be the same, so that the correlation between the target bit and the surrounding bits becomes stronger. As a result, the hit rate of predicting the symbol appearance probability increases, and a high compression effect can be obtained even for character code data.

According to an experiment by the inventor, it was confirmed that the data amount of the character string information of the C language source program is reduced to 40% or less when the encoding apparatus of this embodiment compresses character string information. I have.

In the case of a C language source program,
Although the same character string frequently appears, for example, “0.1258
6 2.13658 5.23698... ", Even in the case of real-valued data in which the same character string hardly appears, the upper bits of each character match, so that the encoding apparatus of the present embodiment Thus, a high data compression effect can be obtained.

In this embodiment, an identification code indicating the data type of an image and a character is inserted at the head of the generated encoded data. This allows the decoding device to determine the data type based on the identification code at the time of decoding, and to automatically execute the respective predetermined operations.

In the above-described embodiment, an example in which character string information of an ASCII code is encoded as code data has been described. However, other character string codes may be encoded, and various non-character code information may also be encoded. can do.
In this case, 1 which is arranged in parallel in the reference symbol acquisition means 3
The number of line bits may be set to one code length.
For example, in the case of JIS code, each character is 2 bytes,
Set the number of bits per line to 2 bytes. Further, the number of bits per line is not limited to one code length, and may be set to an integral multiple of the code length.

Although the identification code inserted into the encoded data indicates only the data type, the identification code may further indicate the number of bits per line.

By the way, the encoded data of the arithmetic code is pseudo-random data, and there is no data pattern which does not appear. For this reason, the specific code cannot be fixedly defined as the identification code. Therefore, for example, first to third various codes are set, and the identification code is defined as a series of the first and second codes, while the first code is added to the encoded data at the time of encoding. Appears, the third code must be inserted afterwards. Thus, upon encoding, the first code can be detected, and when the second code is detected, it can be determined that the code is an identification code.

[0053]

As described above, according to the present invention, since one image coding means encodes both image data and code data, the structure of the apparatus can be simplified and the code data can be simplified. When the data compression is performed, data is arranged in parallel using the number of bits per code data as a unit, and the appearance probability of the symbol of interest is predicted based on the symbol pattern of each bit around the bit of interest. Since the accuracy of the prediction result is high, a high compression effect can be obtained for both the image data and the code data.

[Brief description of the drawings]

FIG. 1 is a block diagram of an encoding device according to an embodiment of the present invention.

FIG. 2 is a block diagram of a decoding device according to an embodiment of the present invention.

FIG. 3 is an operation flowchart of an encoding process.

FIG. 4 is an explanatory diagram showing an operation of arranging input data and extracting each bit data.

FIG. 5 is an explanatory diagram showing a symbol prediction operation.

FIG. 6 is an explanatory diagram showing a method of arranging character code data.

FIG. 7 is an explanatory diagram of output data of an encoding device.

FIG. 8 is an operation flowchart of a decoding process.

[Explanation of symbols]

 DESCRIPTION OF SYMBOLS 1 P / S conversion means 2 Input switching means 3 Reference symbol acquisition means 3a 1 line bit number switching means 4 Probability table 5 Prediction value / probability data selection table 6 Prediction judgment means 7 Arithmetic code generation means 8 Table rewriting means 9 Identification code insertion Means 10 Identification code detection / removal means 11 Arithmetic code decoding means 12 Output switching means 13 S / P conversion means

Claims (4)

(57) [Claims] An image and code data compression method for selectively inputting binary image data in which the number of pixels in one line is constant and code data in which the number of bits in one unit is constant, and compressing the data. A predetermined number of input data are arranged in parallel, a symbol appearance probability of the bit of interest is predicted from the state of each bit around the bit of interest, and data compression is performed by encoding the correspondence between the predicted value and the actual symbol. 1
When two arithmetic coding means are provided and the binary image data is compressed, the number of pixels in one line is set as the fixed number for arranging the data in parallel, and the binary coding is performed by the arithmetic coding means. In the case of compressing the image data while compressing the code data, setting the number of bits of one unit of the code data as the fixed number and compressing the code data by the arithmetic coding means. An image and code data compression method characterized by the following. 2. A method according to claim 1, wherein said code data is a character code. 3. The method according to claim 1, wherein when starting the data compression of the binary image data and the code data, an encoded code is outputted after outputting an identification code indicating a type of the data. Data compression method for the described image and code. 4. The image and code data compression method according to claim 3, wherein said identification code includes information indicating said constant number in which said data is arranged in parallel.