JPH06152986A - Picture compressing method and device - Google Patents

Abstract

PURPOSE:To expand compressed data without deteriorating the picture quality of original picture data by working binarization pseudo half-tone picture data to be compressed so that the run length of a bit can be lengthened only by a logical arithmetic operation. CONSTITUTION:The number of black picture elements included in a picture block P is counted, and the number (b) of the black picture elements corresponding to the block is searched (S1-S2). A preliminarily set reference pattern S is obtained by referring to the number (b) of the black picture elements (S3). Next, an EOR is computed 4 at every element at the respective corresponding position of the picture block P and the reference pattern S (S4). Then, when the results of (m)X(m) are all '0' (S5), the picture block P is completely coincided with the reference pattern P, and a code indicating the reference pattern S is adopted 6 as the code indicating the picture block (S6). When at least one '1' is present within the EOR arithmetic result of the (m)X(m) (S5), the EOR arithmetic results are sequentially read from left to right, and rearranged in one column (S7).

Description

Detailed Description of the Invention

[0001]

BACKGROUND OF THE INVENTION 1. Field of the Invention The present invention relates to a method and apparatus for compressing image data binarized by pseudo halftone processing such as dithering.

[0002]

2. Description of the Related Art It is desired in a storage device such as a facsimile machine or a file system to efficiently compress binary image data obtained by subjecting an original halftone image to pseudo halftone processing such as dithering. As a technique for efficiently compressing binarized image data, CCITT Recommendation T.M. Four
There is an MH / MR encoding method standardized by the. In this encoding method, the run lengths of the black pixels and the white pixels are Huffman-encoded, and short codes are assigned to long run lengths. However, since there are many isolated black pixels / white pixels in the binarized pseudo-halftone image processed by the dither method, it is almost impossible to expect a long run length to occur. Can not be compressed well. As a method for compressing such image data, there is a technique disclosed in Japanese Patent Laid-Open No. 1-141484. This technique divides the binary pseudo-halftone image data to be compressed into 2 × 2 pixel blocks, and the number of black pixels in each pixel block is 2-3.
Bit codes are assigned. In this technology, the amount of data resulting from compression is 1/2 to 3 / of the original amount.
Guaranteed to be 4.

[0003]

However, in the technique disclosed in Japanese Unexamined Patent Publication No. 1-141484, in order to restore the encoded data to the image data, the encoded number of black pixels is set at a predetermined position or not. Since it is arranged intentionally, there is a drawback that the original image data cannot be completely restored.

The present invention has been made in view of the above circumstances, and provides a method and apparatus capable of efficiently compressing binarized pseudo-halftone image data without degrading the image quality during decoding. Is its purpose.

[0005]

An image compression method according to the present invention divides binarized image data into blocks of a predetermined size and corresponds to the number of black images included in an arbitrary image block. The reference pattern to be compared with the image block is compared, and if the image block and the reference pattern match, the code indicating the reference pattern is assigned to the image block. Otherwise, the comparison result is encoded. Is characterized by.

For the image block that does not match the reference pattern, the bit string obtained by holding the previous state for the bit for which the comparison result matches and inverting the previous state for the bit for which the comparison result does not match is encoded. May be. If all the bits of the image block do not match the reference pattern, a code indicating that may be assigned.

The image compression apparatus according to the present invention divides the binarized image data into blocks of a predetermined size and stores the memory, and a reference corresponding to the number of black pixels included in an arbitrary image block. A reference pattern generator that generates a pattern, a comparator that compares the image block with the reference pattern, a pattern number encoder that forms a code representing the reference pattern, and a smoothing process that smoothes the comparison output from the comparator. And a run length encoder that encodes the run length of this smoothed data.

[0008]

According to the present invention, the binary-coded pseudo-halftone image data to be compressed is given a "0" bit only by a logical operation.
Since the run length of '1' bit is processed to be long, the compressed data can be expanded without deteriorating the image quality of the original image data. Further, the MH / MR encoder can be used as it is when encoding the data string obtained by processing.

[0009]

Embodiments will be described below with reference to the drawings.
FIG. 1 is a diagram for explaining the outline of the present invention. Here, for simplification, the size of the image block (dither matrix) is 4 × 4. FIG. 1- (a) shows a state of an arbitrary image block. In the figure, the shaded rectangles represent black pixels. The number of black pixels included in this pixel block is eight. FIG. 1- (b) is a reference pattern for eight black pixels. 1 (a) shows the result of calculating the exclusive OR (EOR) between the image block shown in FIG. 1- (a) and the corresponding pixels of the reference pattern shown in FIG. 1- (b). c). The contents of the EOR block in FIG. 1- (c) are taken out in order from the top and rearranged into one column to form 16-bit data.
It is shown in (d). Here, the portion of the pixel that does not match the reference pattern is "1", which interrupts consecutive runs of "0" and is isolated. FIG. 1- (e) shows the result of smoothing. This process calculates the EOR of the first bit and '0' and replaces the result with the first bit, then calculates the EOR of the second bit and the first bit replaced by this EOR operation result, and outputs the result. Replaced with the second bit,
It continued to 6. As a result, a bit pattern in which the run of "0" or "1" changes at the bit position different from the pixel value of the reference pattern is obtained. As is clear from a comparison between FIG. 1- (d) and FIG. 1- (e), the smoothed pattern has a longer run of “0” or “1” and can eliminate isolated bits. .

At the time of encoding, the smoothed "0" or "1" run of FIG. 1- (e) may be Huffman encoded. An MH / MR code generally used in facsimiles may be used for the encoder at this time.
In addition, when the arrangement of the black pixels of the image block matches the arrangement of the black pixels of the corresponding reference pattern, that is, FIG. 1- (c)
When all the elements of the EOR block of are 0,
A code indicating the reference pattern at that time is assigned without performing the smoothing process.

FIG. 2 is a block diagram showing the configuration of the image compression apparatus of the present invention. In the figure, 1 is m × m in which image data to be compressed is taken out, for example, m pixels every m lines.
Is a block memory that forms and stores the image block P of FIG. A counter 2 counts black pixels included in the image block P. Reference numeral 3 is a reference pattern generator for generating a predetermined reference pattern with reference to the number of black pixels obtained by the counter 2. Reference numeral 4 is a pattern number encoder that forms a code representing a reference pattern. 5
Is a latch for temporarily holding the image data that has been fielded back. A run length counter 6 counts the run length of the image data. A run length encoder 7 performs Huffman coding on the run length obtained by the run length counter 7. Reference numeral 8 is an exclusive OR gate for calculating the EOR of each element at positions corresponding to the image block and the reference pattern. Reference numeral 9 is an exclusive OR gate, which performs a smoothing process by field-backing its output through the latch 5. In this smoothing processing, k of the operation result of the exclusive OR gate 8
For the th element f _k , C _k is obtained by performing the operation shown in Expression 1 described later.

FIG. 3 is a flow chart for explaining an image compression method of the present invention which can be carried out by using the above image compression apparatus. Again, the size of the image block is typically m ×
m. First, m pieces of image data to be compressed are taken out every m lines to form an m × m image block P (S1). Next, this image block P
The number of black pixels included in the block is counted to obtain the black pixel number b for the block (S2). Then, a predetermined reference pattern S is obtained by referring to the black pixel number b (S3). If the dither algorithm that processed the image to be compressed is known in advance, this reference pattern may be determined according to it. Next, EOR is calculated for each element at positions corresponding to the image block P and the reference pattern S (S4).

If all the m × m results are '0' (S
5) Since the image block P at this time completely matches the reference pattern S, the code indicating the reference pattern S at this time is adopted as the code indicating this image block (S6). If at least one "1" is present in the m × m EOR operation results (S5), the EOR operation results are read in order from left to right and rearranged in one column (S7). next,
The k-th element f _k of the rearranged EOR operation result is smoothed by performing the operation shown in Expression 1 to obtain c _k .

C _k = (C _k-1 ) EOR (f _k ) ... (Formula 1) Here, k is a natural number of m × m or less, and the value of C _k-1 when k = 1. Is set to "0" (S8). The image block is Huffman coding the run of the C _K '0' or '1' (S9). When the above processing is performed for all image blocks (S10), the compression of the binary image according to the present invention ends.

FIG. 4 shows an example of the code string obtained in the above description.
This is an example. In FIG. 4, PN is a reference pattern
It means the code that represents the code. Also, HF is "0" young
Means ‘1’-bit Huffman code of run
It Here, each image block P _nFor each
Semi-pattern S_nA code that represents or a smoothing that belongs to a block
The result of processing is 0 or 1 bit run
It is assigned to the Fuman code. At this time, the reference pattern
S_nBlocks that cannot be rewritten with the code that represents
If you want to do that, put those blocks together and
Or you can Huffman-encode a run of '1' bits
Yes.

Here, consider a reference pattern when half of the number of pixels included in an image block is a black pixel. For example, if the size of the image block is 4 × 4, then FIG.
(B) corresponds. If the image block to be compared with the reference pattern is an inverted version of the reference pattern, all comparison results with the reference pattern will be "1". When smoothing processing is applied to this, 0101 ...
.., that is, a run length of "1" and a run length of "0" appear repeatedly, so that Huffman coding produces a code longer than the original data length.
In such a case, the comparison result may be Huffman-encoded as it is, or a code in which all the bits of the image block do not match the reference pattern may be assigned. In the present invention, since the reference pattern is obtained from the number of black pixels included in the image block and compared with the original image block, the number of black pixels and the number of white pixels that do not match the reference pattern are the same. Therefore, in order for the result of the smoothing process to start from '1', the mismatch with the reference pattern must occur at the first pixel, and the mismatch with the reference pattern also occurs at other places. Therefore, the results of the smoothing process are never all '1'.

[0017]

As described above, according to the present invention,
The binary pseudo-halftone image data to be compressed is processed so that the run length of "0" bit and "1" bit becomes long only by logical operation, so that the original image data is not deteriorated in image quality. There is an effect that the compressed data can be expanded.

[Brief description of drawings]

FIG. 1 is a diagram illustrating an outline of the present invention.

FIG. 2 is a block diagram showing a configuration of an image compression apparatus of the present invention.

FIG. 3 is a flowchart for explaining an image compression method of the present invention.

FIG. 4 is a diagram showing an example of a code string generated according to the present invention.

[Explanation of symbols]

 1 Block Memory 2 Counter 3 Reference Pattern Generator 4 Pattern Number Encoder 5 Latch 6 Run Length Counter 7 Run Length Encoder 8, 9 Exclusive OR Gate

Claims (4)

[Claims] 1. The binarized image data is divided into blocks of a predetermined size, and a reference pattern corresponding to the number of black pixels included in an arbitrary image block is compared with the image block. An image compression method characterized by assigning a code indicating the reference pattern to the image block when the image block matches the reference pattern, and encoding the comparison result otherwise. 2. A bit string obtained by retaining the previous state of an image block that does not match the reference pattern with a bit for which the comparison result matches and inverting the previous state for the bit for which the comparison result does not match. The image compression method according to claim 1, wherein the image compression is performed. 3. The image compression method according to claim 1, wherein when all the bits of the image block do not match the reference pattern, a code indicating that is assigned. 4. A memory for dividing binarized image data into blocks of a predetermined size for storage, and a reference pattern generation for generating a reference pattern corresponding to the number of black pixels included in an arbitrary image block. , A comparator for comparing the image block with the reference pattern, a pattern number encoder for forming a code representing the reference pattern, a smoothing processor for smoothing the comparison output from the comparator, and this smoothing device. And a run-length encoder for encoding the run-length of the data.