JP2793536B2 - Image data compression method - Google Patents

Description

DETAILED DESCRIPTION OF THE INVENTION

[0001]

BACKGROUND OF THE INVENTION 1. Field of the Invention The present invention relates to a method for compressing image data, and more particularly to a method for compressing image data for a facsimile
Recommendation T. U.T. 85 by JBIG (Joint B
The present invention relates to a data compression method for compressing data using an i-level image group.

[0002]

2. Description of the Related Art In a facsimile apparatus, general data is compressed and decompressed in order to compress the amount of information to be transmitted and shorten the transmission time. As this compression encoding method, the recommendation T. G.3 facsimile coding (MH, MR system) as described in Recommendation T.4. In addition to the conventional coding method such as the G4 facsimile coding method (MMR method) by ITU-T, the coding method JBIG (facsimile version) was issued by the ITU-T in August 1995. It was officially recommended as 85.

In the conventional MH / MR / MMR, a run length value, which is the length of a continuous pixel of white or black, is obtained from pixel data during encoding processing, and a code corresponding to the run length is encoded in an encoding table. I want more. On the other hand, in the compression processing of one pixel to be coded, JBIG refers to the ten pixels around the pixel to be coded, predicts whether the pixel to be coded is white or black, and if the prediction is actually incorrect. Only the coding is performed by the arithmetic coding method.

[0004] This JBIG has a higher compression ratio than MH / MR / MMR, but has a longer time required for the encoding process itself, and is usually realized by hardware.
Hereinafter, the compression processing by JBIG will be described with reference to the drawings.

FIG. 4 is a block diagram showing a main part of an example of a conventional encoding apparatus for performing a compression process in a facsimile apparatus. This apparatus has a CPU (central processing unit) 10 for performing arithmetic processing, data processing, control processing of each unit, and the like.
0, an image data memory 400 for storing binary pixel data of a document read at the time of transmission, and a code data memory 500 for storing data received at the time of reception as a code.
And an encoding program 201 by JBIG, and an encoding / decoding table 20 for converting image data into encoded data or converting encoded data into image data.
2, a learning table memory 300 for predicting whether a pixel to be coded is white or black and increasing the probability of hitting it, and code data or image data as a result of the conversion. (First-in first-out) and a FIFO memory 600 that is managed by a first-in first-out (FIFO).

The procedure of the compression process shown in FIG. 4 will be described with reference to the flowchart of FIG. 5 and the memory arrangement diagram of the image data of FIG. First, original data read by a scanner or the like is converted into binary pixel data and stored in the image data memory 400. This image data compression processing is sequentially performed from the upper left to the lower right of the document.

The compression process for one line of image data will be described with reference to FIG. The image data L1 includes an encoding target line L11, a line L12 immediately before the encoding target line L11,
Each of the lines L13
11 to L13 are stored in the memory blocks B1 to B3. For the JBIG, whether to perform general-purpose prediction or not (step S11) can be designated by a JBIG free parameter TPBON. Here, general-purpose prediction is
The line to be encoded (line L1) and the line immediately before the line to be encoded (line L12) are compared (step S12), and it is determined whether these two lines are the same (step S13).

If the result of determination in step S13 is not the same, one-line encoding processing (step S14)
And if they are the same, the encoding target line L11 is not set as the encoding target. Therefore, by performing general-purpose prediction, the number of lines to be encoded can be reduced, and the compression time can be reduced.
When the general-purpose prediction is not performed, the one-line encoding process (step S15) is performed as it is.

Next, one-line encoding processing S14 (S15)
Will be described with reference to the data arrangement diagrams of FIGS. 7 and 8 and the flowchart of FIG. Here, the case where the JBIG free parameter LRLTWO is OFF will be described. JB
The IG-free parameter LRLTWO is a group of 10 pixels to be referred to when predicting a pixel to be encoded. When the parameter LRLTWO is OFF, as shown in FIG. 7A, the reference 10 pixels extend over 3 lines. And is called a three-line model template. Also, the parameter LR
When LTWO is ON, as shown in FIG. 8, the reference 10 pixels extend over two lines and are called a two-line model template.

First, a block unit (block B) from the line (line L13) immediately before the line L11 to be encoded is described.
The pixel data is read out in 3) (step S21). Here, the block unit is 32 pixels. Next, image data is read out in blocks (block B2) from the line (line L12) immediately before the line to be encoded (step S22). Further, the encoding target line (line L11)
Image data is read out in block units (block B1) (step S23). After reading the memory blocks for three lines, the reference 10 pixels are extracted from the three memory blocks B3 to B1 at the pixel positions shown in FIG. 7A, and the context X is created (step S2).
4).

Here, the context X is a one-dimensional representation of 10 pixels around the reference. One-pixel encoding processing (step S25) is performed using the context X.
Specifically, the context is used to predict whether the encoding target pixel is white or black. The prediction error is regarded as a symbol sequence, and the symbol sequence is compressed by arithmetic coding. At the same time, the probability of prediction hit or miss is calculated for each context value and stored in the learning table memory. By using the learning table, it is possible to increase the probability that the prediction is successful and to reduce the size of the compressed code data.

After the one-pixel encoding process, it is determined whether or not the compression has been completed for the pixels of the block unit (step S26). If the compression has not been completed, the memory blocks B3 to B1 are determined.
Is shifted by one bit (steps S27 to S29), and the process returns to the context creation processing (step S24). By this one-bit shift processing, as shown in FIG. 7B, the reference 10 pixel position for compressing the next encoding target pixel can be determined. When the processing for the block unit is completed, the process returns to the beginning of the one-line encoding processing, and it is determined whether the encoding processing for one line is completed (step S2).
0). The compression process is performed by repeating the above process for the number of lines of the document.

[0013]

A first problem of the above-mentioned conventional compression method is that the compression time is MH / MR / MM.
This method is longer than the compression processing using R. The reason is J
The compression processing of image data by the BIG is basically performed in units of one pixel, and when general-purpose prediction is performed, that is, even when two adjacent lines are the same, the next line of the two adjacent lines is copied. This is because the memory block for three lines is unconditionally read without referring to the general-purpose prediction result when encoding.

The second problem is that the line to be coded is
It is necessary to unconditionally store image data for a total of three lines, one line before the line to be encoded and two lines before the line to be encoded, until the compression processing of the line to be encoded is completed. There is poor memory efficiency. The reason is that the memory block for three lines is unconditionally referred to in the context creation processing.

An object of the present invention is to solve these two problems, and in particular, to provide an image data compression method in which binary image data compression processing by JBIG is realized by software.

[0016]

According to the structure of the present invention, image data is read from an image memory in units of blocks for each line, and a context is created for each pattern of the image data in units of blocks to encode the image data. In an image data compression method for compressing binary pixel data with reference to peripheral pixels of a target pixel, the context creation processing includes:
It is characterized in that it is determined whether the image data in the block unit is all white or other than all white, and if this is all white, it is not set as context creation target data.

In the present invention, when image data is read from the image memory, the data of the line two lines before the line to be coded in the image memory and the data of the line immediately before the line to be coded are read.
Comparing the chromatography data, it reads as image data only one line of image data of the two lines if they match, it is also possible to not read the image data of the other line, further coded Two lines before the line
Compare the data of the line with the data of the previous line ,
If they do not match, the image data is read from the image memory in units of blocks for each line, and context creation processing is performed for each pattern of the image data in units of blocks, and the image data in units of blocks is all white or all white. It is also possible to determine whether the data is other than the context creation target data when this is all white.

According to the present invention, in the process of reading out image data from the image memory in units of blocks for each line and creating a context for each pattern, it is determined whether the image data is all white or other than all white. In the case of a template, it is divided into eight patterns. Here, since the all white lines are not used as context creation target data, mask processing and shift processing of a memory block can be omitted, and the compression processing time can be shortened accordingly.

Also, a line two lines before the line to be encoded
Is compared with the data of the previous line, and if they are the same, the encoding target line is excluded from the encoding target. Two lines before
With reference to the determination result of the match / mismatch between the data of the line and the data of the previous line, if they match, only one of the two lines of image data is read out, and the image data of the other line is read out. By not reading,
Since the compression processing time is shortened and it is not necessary to hold the content of the image data of one line, the memory efficiency can be improved.

[0020]

Embodiments of the present invention will be described below with reference to the drawings. FIG. 1 is a block diagram for explaining a first embodiment of the present invention, and FIG. 2 is a flowchart showing an encoding processing procedure.

As shown in FIG. 1, this embodiment employs arithmetic processing,
CPU (central processing unit) 100A that performs data processing and control processing of each unit, a code data memory 500 that stores binary image data of a document read at the time of transmission, and an encoding program 201A by JBIG
And a ROM 200A in which a coding / decoding table 202 for converting image data into code data or converting code data into image data is arranged and stored, and predicting whether the pixel to be coded is white or black. A learning table memory 300 referred to in order to increase the probability and code data or image data as a result of the conversion are stored in the FI
F managed by FO (first in first out)
The IFO memory 600 is connected by a bus 700.

The CPU 100A, which is a feature of the present invention, internally stores image data A in units of blocks from a line to be encoded.
And a register 104 for holding data obtained by reading image data B in block units from the line immediately before the encoding target line.
And a register 105 for holding data obtained by reading image data C in block units from a line two lines before the line to be encoded, and a register 1 for holding context contents
02 and a register 101 for holding a general-purpose prediction result. The CPU 100A executes a JBIG encoding program 201A described below.

Next, the procedure of the JBIG encoding program 201A of this embodiment will be described with reference to the flowchart of FIG. Here, similarly to the conventional example, a case where the JBIG free parameter LRLTWO is OFF will be described. First, the line immediately before the line to be encoded (line L
13) The image data is read out in block units (block B3) (step S21). This block unit is also 3
Two pixels. Next, image data is read out in blocks (B2) from the line (line L12) immediately before the line to be encoded (step S22). Further, the encoding target line (line L1) is used in block units (block B).
The image data is read in 1) (step S23).

The three memory blocks B read here
It is determined whether 3-B1 is all white (step S3).
0), if the result of this determination is not all white, the reference 10 pixels are extracted from the three memory blocks B3 to B1 at the pixel positions shown in FIG. 7A and the context X is created as in the conventional case (step S24). ). Here, the context is a one-dimensional representation of 10 pixels around the reference. One-pixel encoding processing (step S25) is performed using the context X.

Specifically, the context is used to predict whether the pixel to be encoded is white or black. The prediction error is regarded as a symbol sequence, and the symbol sequence is compressed by arithmetic coding. At the same time, the probability of prediction hit or miss is calculated for each context value and stored in the learning table memory. By using this learning table, it is possible to increase the probability that the prediction is successful and to reduce the size of the compressed code data.

After the one-pixel encoding process (S25), it is determined whether or not the compression has been completed for the pixels of the block unit (step S26). If not completed, the memory blocks B3 to B1 are shifted by one bit. (Step S27
To 29), and return to the context creation processing (step S24). By this one-bit shift processing, FIG.
As shown in (b), the reference 10 pixel position for compressing the next encoding target pixel can be determined. When the processing for the block unit is completed, the process returns to the top of the one-line encoding processing, and it is determined whether the encoding processing for one line is completed (step S20).

Further, in step S30, the read
If the three memory blocks B3 to B1 are all white, the block processing ends until the block processing ends (step S32).
No context creation processing is performed, and only one-pixel encoding processing (step S31) is performed. The reason for this is that when the three memory blocks B to B1 are all white, the context remains unchanged at 0 during the block unit processing. Further, there is no need to shift the memory blocks B3 to B1 by 1 bit after the one-pixel encoding process (step S32). Thereby, three memory blocks B3
As in the case of an image in which .about.B1 is all white, that is, a document image which is often handled by facsimile, the number of times of context creation processing can be greatly reduced with respect to the compression processing of an image that wins with white eyes. Even when the three memory blocks are not all white, for example, the memory block B1 of the encoding target line (line L11) is white,
For example, when the other two memory blocks are not white, by performing a dedicated context creation process for each of the three memory block patterns, the compression processing speed can be further improved.

Next, the JBIG according to the second embodiment of the present invention will be described.
Will be described with reference to the flowchart of FIG. Here, as in the conventional example, JBIG
The case where the free parameter LRLTWO is OFF will be described. First, a comparison result between the line immediately before the encoding target line (line L13) and the line immediately before the encoding target line (line L12) is referred to (step S4).
0). Note that this comparison reference result is stored in the register general-purpose prediction determination result 101 in the CPU 100A. With reference to this determination result, it is determined whether the two lines are the same (step S41). If they are not the same line, the processes from step S20 to step S29 are performed for one line as in the conventional example. As a result of the determination, if the line (line L13) immediately before the line to be encoded is the same as the line (line L12) immediately before the line to be encoded, the line before the line to be encoded is the same. line (line L12) reads the image data in from the blocks (block B2) (step S22A), further encoding target line (line L11) from each block (block B
The image data is read in 1) (step S23A).

With respect to the line (line L13) two lines before the line to be coded, image data is not read out in block units, and the line (line L13) immediately before the line to be coded already read out in step S22A is read. Line L1
The context is created from the memory block (block B1) read out from the encoding target line by referring to the image data in block unit of 2) (step S42). One-pixel encoding processing (step S25A) is performed using this context.

After the one-pixel encoding process, it is determined whether or not the compression has been completed for the pixels in the block unit (step S26A).
2 and B1 are shifted by one bit (steps S28A, 2
9A), the context creation processing (step S42)
Return to before. By this one-bit shift processing, FIG.
As shown in (b), the reference 10 pixel position for compressing the next encoding target pixel can be determined.

Here, it is not necessary to perform the one-bit shift process on the memory block of the line (line L13) two lines before the line to be encoded. Further, the line (line L13) immediately before the line to be encoded and the line (line L12) immediately before the line to be encoded are the same, so the line (line L13) immediately before the line to be encoded is the same. ) Need not be stored in the image data memory 400 and can be released as a free memory area. When the processing for the block unit is completed, the process returns to the top of the one-line encoding processing, and it is determined whether the encoding processing for one line is completed (step S20A). By referring to the general-purpose prediction result, the number of times of reading image data can be reduced, and the memory efficiency can be increased.

[0032]

As described above, according to the present invention,
A two-image data compression method that refers to the peripheral pixels of the encoding target pixel, or one of the encoding target line and the encoding target line
The previous line is compared, and if they are the same, the encoding target line is excluded from the encoding target, so that the encoding processing speed can be improved. This is due to a reduction in the number of context creation processes and a reduction in the number of times image memory is read.

According to the present invention, the line to be coded is compared with the line immediately before the line to be coded, and if they are the same, the line to be coded is excluded from the lines to be coded.
Since the unnecessary image memory is released by using the general-purpose prediction result, the memory efficiency can be improved.

[Brief description of the drawings]

FIG. 1 is a block diagram showing a main part of a first embodiment of the present invention.

FIG. 2 is a flowchart showing an encoding procedure of FIG. 1;

FIG. 3 is a flowchart showing an encoding procedure according to the second embodiment of the present invention.

FIG. 4 is a block diagram showing a main part of a portion for performing image data compression in a conventional example.

FIG. 5 shows a conventional binary image data compression method having means for comparing a line to be coded with a line immediately before the line to be coded and, if they are the same, excluding the run-in to be coded from the lines to be coded. 3 is an overall flowchart of FIG.

FIG. 6 is a data arrangement diagram for explaining the relationship between encoding target image data, lines, and memory blocks.

FIG. 7 shows a memory block of a line to be coded and 3
FIG. 8 is a data arrangement diagram for explaining a relationship between a line model template and a context.

FIG. 8 shows a memory block of a line to be coded and 2
FIG. 9 is a data arrangement diagram illustrating a relationship between line models and templates.

FIG. 9 is a flowchart showing a conventional encoding procedure.

[Explanation of symbols]

100, 100A CPU 101 General-purpose prediction result register 102 Context register 103-105 Memory block register 200, 200A Program and table ROM 201, 201A JBIG coding program 202 Coding / decoding table 300 Learning table memory 400 Image data Memory 500 code data memory 600 FIFO memory 700 bus L1 image data L11 line to be encoded L12 line immediately before the line to be encoded L13 line immediately before the line to be encoded B1, B4 line to be encoded Memory block B2, B5 Memory block of the line immediately before the line to be encoded B3, B6 Memory block of the line before the line to be encoded

Claims (3)

(57) [Claims] An image data is read out from an image memory in units of blocks for each line, a context is created for each pattern of the image data in units of blocks, and reference is made to pixels surrounding the pixel to be encoded of the image data. In the image data compression method of compressing the value pixel data, the context creation processing determines whether the image data in the block unit is all white or other than all white. A method for compressing image data, comprising: 2. When image data is read from an image memory, a line immediately before an encoding target line of the image memory is read.
The data of one line is compared with the data of the previous line, and when they match, only one of the two lines of image data is read as image data , and the image data of the other line is not read. An image data compression method, characterized in that: 3. The data of a line two lines before the line to be encoded.
The data is compared with the data of the immediately preceding line, and if they do not match, the image data is read from the image memory in units of blocks for each line, and the context creation processing is performed for each pattern of image data in units of blocks. 3. The image data compression method according to claim 2, further comprising: determining whether the image data in the block unit is all white or other than all white.