JPH10108013A - Image compression device - Google Patents

Abstract

PROBLEM TO BE SOLVED: To place a main purpose on the isolated point elimination and notch elimination of images, to improve compressibility and the image quality of and to perform a high-speed processing several hundreds times faster than previously by performing pattern matching to run length data. SOLUTION: Image pixel data area inputted from an image input part (S201) and a pixel data → run length data conversion processing is performed (S203). Then, a notch pattern and an isolated point pattern in the run length data are detected (S204), and the conversion processing of the run length data is performed for the detected patterns (S205). For the conversion-processed run length data, conversion processing into compressed data of MR and MMR codes or the like is performed (S206). The image conversion processings are performed until the end of image data (S202), and after the conversion processing are ends, the compression data are transferred to an external information equipment form an image communication IF part or are stored in an image storage part (S207).

Description

DETAILED DESCRIPTION OF THE INVENTION

[0001]

BACKGROUND OF THE INVENTION 1. Field of the Invention The present invention relates to an image compression / decompression apparatus, and more particularly to a CCI used in a facsimile apparatus.
The present invention relates to an apparatus for improving a compression ratio when an image is compressed by TT standard MR encoding and MMR encoding.

[0002]

2. Description of the Related Art In recent years, image processing techniques such as facsimile machines, digital copiers, desktop publishing machines, and image filing machines have been applied to many OA products. In these devices, image data is often compressed and stored or transmitted due to its large capacity.

As a typical image compression method, a MH (Modified) standard of CCITT (International Telegraph and Telephone Consultative Committee) international standard used in facsimile machines is used.
Huffman method, MR (Modified RE)
AD) system and MMR (Modified MR) system. These systems and their realizing devices are disclosed in
Details are described in, for example, JP-A-126368. MH
The encoding method uses only one-dimensional image signal correlation,
The MR and MMR coding schemes are more efficient coding schemes utilizing two-dimensional correlation.

[0004] The MH, MR and MMR encoding methods encode an original image faithfully without any deterioration.
For these, a method has been proposed in which image processing conversion is performed before encoding to reduce the code amount. "Reduction of information amount of facsimile signal and image quality evaluation by thinning processing" (Transactions of the Institute of Electronics, Information and Communication Engineers, D-II Vol. J73-DI)
I No. 9 pp. 1538-1544 (September 1990), the compression code is reduced by performing a thinning process on the original image before encoding, and a thickening process at the time of decoding.

In Japanese Patent Laid-Open No. Hei 6-268842, a reduction in the number of compression codes is realized by extracting and changing a specific pattern from an original image during MR and MMR coding. Furthermore, "Notch removal method using variable window" (1990 IEICE Spring National Convention D-
In 425), the image quality and the compression ratio are improved by removing the notch pattern using the variable window.

[0006]

MH, MR and MMR
Various techniques for improving the compression ratio have been proposed for the encoding method as described above. However, since these methods perform pattern matching and image conversion on an original image before encoding, there is a problem that an extremely long processing time is required. For example, according to the measurement results of the inventors,
As a result of investigating the pattern of noise elimination and notch elimination in Japanese Patent No. 268842 by CPU software processing, a processing time on the order of several minutes was required for an A4, 200 DPI image.

In Japanese Patent Application Laid-Open No. 6-268842, since a pattern is extracted for the purpose of improving the compression ratio, pattern matching with a very large number of pixels in addition to the above pattern is also presented. In addition, the processing time is expected to increase several times, and there is a problem that image quality degradation is not considered. In the above-mentioned other known examples, it is necessary to perform pattern matching with a large number of pixels.
There is a problem that a processing time equivalent to or longer than that of Japanese Patent Publication No. 2 is required.

An object of the present invention is to perform pattern matching on run-length data for MR and MMR compression methods, which are two-dimensional compression methods, so that about 1/100 of the conventional method without using special hardware. Alternatively, high-speed compression can be achieved in a processing time shorter than that to realize a speed in which the load on CPU processing can be ignored, and MR, MMR
An object of the present invention is to provide means for improving the compression rate and the quality of an image while focusing on the removal of isolated points and the notch of an image while following the compression procedure of the compression method.

[0009]

In order to achieve the above object, an image compression apparatus according to the present invention comprises: an image input means for inputting image data; and a pixel data for converting pixel data of an input image into run-length data. Run-length data conversion means, run-length data for extracting and converting a notch pattern → run-length data high-compression data conversion means, and a run for converting the converted run-length data into compressed data such as MR and MMR. Length data → compressed data conversion means.

In the present invention, pixel data of an image is represented by M
In the image compression processing for converting to compressed data such as R, MMR, etc., run-length data is introduced as intermediate generation data, and the notch removal of the image is performed on the data, so that compressed data such as MR, MMR, etc. In addition to realizing high-speed compression data reduction, image quality is improved by removing notches.

[0011]

DETAILED DESCRIPTION OF THE PREFERRED EMBODIMENTS One embodiment of the present invention will now be described with reference to FIG.
This will be described below.

In FIG. 1, reference numeral 101 denotes an image input unit for receiving image data from a scanner, a keyboard, a mouse, an external information device, or the like; 102, pixel data for converting input pixel data into run-length data → run-length data A conversion processing unit 103 extracts an isolated point pattern or a notch pattern from the run-length data and converts the pattern into run-length data → run-length data high-compression data conversion processing unit 104
Is a run-length data-to-compressed data conversion processing unit that converts the run-length data that has been subjected to the high-compression data conversion processing to compressed data such as MR and MMR codes;
Reference numeral 105 denotes compressed data for expanding data such as MR and MMR codes which have been subjected to compression processing to pixel data → pixel data restoration processing unit. Reference numeral 106 denotes an image communication IF for transferring data of compressed data and pixel data to and from an external information device. 107, an image storage unit for storing compressed data or pixel data; 108, an image output unit for outputting the expanded pixel data to a CRT, a printer, etc .; and 109, mutually exchanging data. Image / data bus.

Next, with regard to the image high compression conversion / image compression processing in the embodiment shown in FIG. 1, a problem analysis diagram of the flowcharts shown in FIGS.
This will be described in detail using an analysis diagram (hereinafter, referred to as a PAD diagram).

The process of converting an image to be processed to compressed image data will be described in detail with reference to FIG. First, from the image input unit 101,
Image pixel data is input (S201). The input image data may be data read from another information device via the network and the image communication IF unit 106, input from the image input unit 101 using a scanner, a keyboard, a mouse, or the like. The image stored in the storage unit 107 may be read.

The image data to be input may be configured such that the original image pixel data is directly input from the image input unit 101. Alternatively, the image data may be in a format compressed using some compression technique, and may be expanded by the image input unit 101 or the pixel data → run-length data conversion processing unit 102. The compression format is not specified. Further, the compression may not be completely lossless.

Next, image height compression conversion processing is performed from the beginning of the image data until the end of the image data (S202).

In the image high compression conversion processing, first, pixel data → run length data conversion processing is performed on input image pixel data (S203).

Next, the run-length data → run-length data high compression data conversion processing unit 103 first detects a notch pattern and an isolated point pattern in the run-length data (S204). The run-length data-to-run-length data high-compression data conversion processing unit 103 performs a run-length data conversion process on the detected pattern (S205).

Further, the run-length data → compressed data conversion processing unit 104 converts the converted run-length data into compressed data such as MR and MMR codes (S206). The conversion process to the compressed data such as the MR and MMR codes is performed in the same manner as the original image data.

The above image conversion processing is performed until the end of the image data (S202).

After the completion of the conversion process, the compressed data is transferred from the image communication IF unit 106 to an external information device (S20).
7). Alternatively, the image data is stored in the image storage unit 107 (S20).
7). As the image storage unit 107, a configuration using a magnetic disk device or a magnetic disk library device, a configuration using an optical disk device or an optical disk library device, a configuration using a magnetic tape device or a magnetic tape library device, a CD-ROM device or a CD-ROM
A configuration using a library device or the like is applied. Further, a configuration in which these devices are combined is also applied.

The process of converting the compressed image data to the pixel image will be described in detail with reference to FIG. First, convert the compressed image to compressed data →
Input to the pixel data restoration processing unit 105 (S30
1). The input compressed image data is read from another information device via the network and the image communication IF unit 106, the image stored in the image storage unit 107 is read, and the run-length data → compressed data conversion processing unit 104 Configuration for transferring compressed image, image input unit 101
A configuration for inputting compressed image data is applied.

Next, image data restoration processing is performed from the beginning of the compressed image data until the end of the compressed image data (S302).

The image data decompression processing is performed by the compressed data → pixel data decompression
The compression data of the R and MMR codes is restored to the pixel data of the dot image (S303). This image restoration processing is performed until the end of the compressed image data (S30).
2).

After the restoration processing is completed, the pixel image data is output to a display / printer or the like using the image output unit 108. Alternatively, the image data is stored in the image storage unit 107 or output from the image communication IF unit 106 to an external information device (S3).
04).

FIG. 4 shows an example of run-length data.
The run-length data (line table) focuses on the change point of the pixel color (black / white) in the image bitmap sequence, and the change point position of the black / white pixel (or the length of the black / white pixel) for each line. Is shown in a table. Also,
In this example, the start address of the table is set to address “0”.

In FIG. 4, the left end virtual change point pixel of each line is coordinate 0 (the actual left end pixel is coordinate 1), and the pixel color change point traced from the left side of each line is a table. Assume length data.

In the run length data (line table) shown in FIG. 4, pixel color change points are held as a table. Also, for each line, even addresses in the table represent points of change to white pixels, and odd addresses represent points of black pixel transition.

Hereinafter, a first method in a high-speed notch removal method using run length will be described. 16 to 18,
1 shows a first example of a pattern to be removed. 16 to 18 show the following pattern examples.

(1) For patterns [0] to [9],
Removal of isolated points.

(2) One pixel notch removal for patterns [10] to [25].

(3) Two-pixel notch removal for patterns [26] to [37].

(4) Removal of three pixel notches for patterns [38] to [49].

(5) Cornering for patterns [50] to [53].

In this pattern example, the pixels with white circles are the pixels to be removed (that is, the pixel colors are inverted). In addition, although the pattern example of the main removal shows the pattern of the removal of the black pixel, the removal pattern of the white pixel and the black pixel in all the patterns, that is, the removal of the white pixel is also included.

FIG. 5 shows an example of the isolated point pattern removal.
Here, removal of an isolated point of three pixels of the pattern 3 will be exemplified.

In this example, 39, 40 pixels of n lines, n
39 pixels on the +1 line are isolated points. At this time, (1) First, the pixel change point of the nth line is examined. On the nth line, the pixel 39 of i1 is given as a black change point from the line table. The pixel 41 is given as a white change point from the next line of the line table, that is, the next change point i2. Therefore, the pixels 39 and 40 on the n-th line satisfy the conditions of the removal patterns in FIGS.

(2) Next, the (n-1) th line is examined. n-
One line has a white change point of pixel 38 from the line table.
The black change point is the pixel 44. Therefore, the pixel 38 of h1
The pixel 41 of h2 is a white pixel, and the pixel 39 of the n-th line,
Reference numeral 40 corresponds to the conditions of the removal patterns shown in FIGS.

(3) Next, the (n + 1) th line is examined. n +
In one line, the pixel 39 of j1 is given as a black transition point from the line table. The pixel 40 is given as a white change point from the next line of the line table, that is, the next change point j2. Therefore, the pixels 39 on the (n + 1) th line are shown in FIGS.
8 are satisfied.

(4) Finally, the n + 2 lines are examined. n
The +2 line has a white transition point of pixel 3 from the line table.
5. The black changing point is the pixel 42. Therefore, pixel 3 of k1
Pixels 8 to k2 are white pixels, and this pattern is shown in FIG.
It is found that this applies to the removal of the isolated point of the pattern [3] of No. 6.

(5) At this time, the pixels 39 and 4 on the n-th line
The run length table is changed for the pixels 39 on the 0, n + 1 lines. That is, Address1 of the n-th line
Erase data of Nos. 7 and 18, and further add n + 1 line
The data of the ress 11 and 12 are erased.

For other patterns, pixel change points on the nth line are extracted, and pixel patterns on lines above and below the change point are examined using a run-length table to remove black pixel notches and white pixels. The notch is removed.

In the above-described notch or isolated point removing method according to the first method, high-speed processing is possible because pattern matching is performed only on pixel color change points of each line by using a run-length table. Become.
As a result of verifying the CCITT standard manuscript and the A4 image, the number of pixel color change points per line and about 1700 pixels is about 26 change points on average, which is about 1/60 to 1/70 of the actual number of image pixels. It can be executed by pattern matching. Further, since the pattern to be removed is executed only for the pattern of the isolated point, the notch pixel and the missing corner, it is possible to obtain high quality image.

However, in the removal method according to the first method, it is necessary to perform matching on run-length data of up to five lines in the case of a three-pixel pattern in order to verify the removal pattern, and the effect expected in terms of speed is obtained. In some cases, it could not be obtained. Therefore, a high-speed / high-efficiency notch or isolated point removal method for verifying only three lines is proposed as a notch or isolated point removal method according to the second method.

Hereinafter, a second method of a high-speed notch removal method using run length (including an isolated point or a corner loss outside the notch, and hereinafter, typically referred to as a notch) will be described. FIG. 6 shows a second example of the notch pattern to be removed. In FIG. 6, the upward and downward notches are indicated by an image bitmap sequence. In FIG. 6, notches of 1 to 3 pixels are considered as objects to be removed, but there is no particular limitation. The pixel indicated by “Don't Care” in FIG. 6 may be either a black pixel or a white pixel.

Hereinafter, the notch shown in FIG.
The removal method using run-length data will be described.

In the run-length data (line table) shown in FIG. 4, pixel color change points are held as a table. Also, for each line, even addresses in the table represent points of change to white pixels, and odd addresses represent points of transition to black pixels.

Therefore, the data at address 2k + 1 [2k +
1] (k is an integer of 0 or more), the data at address 2k [2
k] is given as the k-th white run length. This is represented by the following equation.

[2k + 1] − [2k]... White run length (Equation 1) (Note: Only the first white run length is derived by 1)
Further, a value obtained by subtracting the data [2k + 1] at the address 2k + 1 from the data [2k + 2] at the address 2k + 2 (k is an integer of 0 or more) is given as the k-th black run length. This is represented by the following equation.

[2k + 2] − [2k + 1]... Black run length (Equation 2) Next, the notch determination processing will be described.

FIG. 7 shows the definition of the line from which the notch is removed and the upper and lower lines. As shown in FIG. 7, the line to be processed is n lines, the upper line is n-1 line, and the lower line is n + 1 line.

First, a search for a black notch is performed for the line to be processed. The k-th black run length is determined for the n-line run-length table. When the black run length given by (Equation 2) is equal to or less than the threshold value (here, 3), the black notch condition is determined.

FIG. 7 shows a black upward notch. The black notch coordinates are set as i1 and i2. Next, a search is made for a white transition point h2 ahead and a white transition point h2 'behind the n1 line with the i1 coordinate in between. This is represented by the following equation.

[H2] <[i1] ≦ [h2 ′] where h2 and h2 ′: continuous white change points (Equation 3) At this time, as shown in FIG. 7, the next black change point h1 ′ after h2.
Are located behind the i2 coordinate, i1 and i2 are defined as black upward notches. This is represented by the following equation.

[H2] <[i1] and [i2] <[h1 ′]... Black upward notch condition (Equation 4) FIG. 8 shows a white notch determination method. As shown in FIG. 8, the line to be processed is n lines, the upper line is n-1 line, and the lower line is n + 1 line.

First, a search for a white notch is performed for the line to be processed. The k-th white run length is determined for the n-line run-length table. When the white run length given by (Equation 1) is equal to or smaller than the threshold value (here, 3), the white notch condition is determined.

FIG. 8 shows a white upward notch. The white notch coordinates are set as i1 and i2. Next, a search is made for a black transition point h2 ahead and a black transition point h2 'behind the i1 coordinate on the n-1 line. This is represented by the following equation.

[H2] <[i1] ≦ [h2 ′] where h2 and h2 ′: continuous black change points (Equation 5) At this time, as shown in FIG. 8, the next black change point h1 ′ after h2.
Are located behind the i2 coordinate, i1 and i2 are defined as white upward notches. This is represented by the following equation.

[H2] <[i1] and [i2] <[h1 ′]... White upward notch condition (Equation 6) FIG. 9 shows a black lower notch determination method. As shown in FIG. 9, the line to be processed is n lines, and the line above it is n−
One line and a line below the line are referred to as an (n + 1) th line.

First, a search for a black notch is performed for the line to be processed. The k-th black run length is determined for the n-line run-length table. When the black run length given by (Equation 2) is equal to or less than the threshold value (here, 3), the black notch condition is determined.

The black notch coordinates are set as i1 and i2. next,
In the (n + 1) th line, a black transition point k2 ahead and a black transition point k2 ′ behind the i1 coordinate are searched. This is represented by the following equation.

[K2] <[i1] ≦ [k2 ′] where k2 and k2 ′: continuous black change points (Equation 7) At this time, as shown in FIG. 9, the next black change point k1 ′ after k2.
Are located behind the i2 coordinate, i1 and i2 are defined as black notches. This is represented by the following equation.

[K2] <[i1] and [i2] <[k1 ′]... Black Notch Notch Condition (Equation 8) For both black / white notches, either one of the upward notch, the downward notch, or both conditions are satisfied. If it does, the notch is removed.

FIG. 10 shows the notch removal processing.
The notch removal processing is realized by deleting the data of the run length table.

FIG. 10 shows the correspondence between the processing for removing the notches i1 and i2 in the n-th line of pixel data and the processing for removing the notches in the run-length table. The n-line original image run-length tables (Address 17, 18) corresponding to the notches i1, i2 in the pixel data are deleted, and n
Generate a line high compression image run length table. Although only the black notch removal is described here, the same processing is performed for the white notch removal.

As the notch removal method, a method of performing only black notch or white notch removal processing, or a method of performing both black notch / white notch removal is considered. As for the notch determination, a method of performing the notch determination only on the original image, a method of sequentially converting the upper line to a corrected image to determine the notch, a method of repeatedly performing the notch processing, and the like are considered.

FIG. 11 shows a method of performing the notch determination only on the original image (not on the deleted image). In this method, the notch determination based on the corrected run-length data is not specifically mentioned, but this does not necessarily mean that the notch is removed only from the original image. It does not exclude notch removal.

FIG. 12 is a flowchart showing the black notch removal for each line. In the black notch removal processing, the original image line table of three lines (S1309, S1310, S13
1311) and a black notch removal line table holding the pointer and black notch removal data (S1312)
And its pointer is needed.

First, each line table is initialized (S1301). 0 is assigned to address 0 of the black notch removal line table. Next, a line end determination process is performed (S1302). If it is determined that the end of the line has not been reached, the process of deriving the black run length is performed next (S130).
3). At this time, the length of the black run length is determined, and if it is longer than the determined notch length, the notch removal processing is not performed (S
1304).

Next, when the length of the black run corresponds to the notch length, the pixel color of the upper line is determined first, and the black notch condition is determined (S1305). Further, the pixel color of the lower line is determined, and the black notch condition is determined (S130).
6). If a black notch is determined in the determination processing in (S1305) and (S1306), black notch removal processing, that is, non-output processing to the black notch removal line table is performed (S1307). If the above condition for black notch removal is not satisfied, black run preservation processing is performed (S1308).

FIG. 13 shows a white notch removal flow for each line. In the white notch removal processing, the black notch removal line table (S1410) and its pointer, the original image line table of the previous line (S1409) and the next line (S1411) and its pointer, and the high-compression image data after conversion are held. Highly compressed image line table (S14
12) and its pointer is needed.

First, each line table is initialized (S1401). 0 is assigned to address 0 of the high-compression image line table. Next, a line end determination process is performed (S1402). If it is determined that the line is not at the end, a process of deriving a white run length is performed (S1403).
At this time, the length of the white run length is determined, and if not longer than the determined notch length, the notch removal processing is not performed (S140).
4).

Next, when the length of the white run length corresponds to the notch length, first, the pixel color of the upper line is determined, and the white notch condition is determined (S1405). Further, the pixel color of the lower line is determined, and the white notch condition is determined (S140).
6). If it is determined in the determination processing in (S1405) and (S1406) that a white notch has occurred, white notch removal processing, that is, non-output processing to the high-compression image line table is performed (S1407). If the above-mentioned condition of white notch removal is not satisfied, white run preservation processing is performed (S1408).

In the processing shown in FIGS. 12 and 13, the black notch / white notch removal is separately executed for each line. Next, a black / white notch removal one-pass processing method for each line will be described.

FIG. 14 shows an example of the one-pass processing of black / white notch removal. The notch is determined sequentially from the left end of each line.
When the black notch is removed, the black run on the right side of the next white run becomes white, so the white run is not subjected to the removal processing. Similarly, when the white notch is removed, the black run is continued for the next right black run, so that the removal processing is not performed.

FIG. 15 shows the flow of the one-pass processing for removing the black / white notch. Basically, the black notch removal processing (S160
2) and white notch removal processing (S1603).
Here, the one-pass processing does not mean that the black notch processing and the white notch processing are separately performed twice for each line as shown in FIG. And white notch processing are performed in parallel. That is, in the n-line matching shown in FIG. 14, after performing the third to fifth black notch removal processing, the black notch removal flow is continued, the eighth black notch removal processing is performed, and then the 10th to 13th black notch removal processing is performed. After storing the notches, the processing shifts to the white notch processing flow to remove 14 to 15 white notches.

First, initialization is performed (S1601).
First, the notch removal process is a black notch determination / removal process (S1).
602). When the black notch removal processing is performed in (S1602), the black notch removal flow (S1602)
1602) is performed again. If the black notch removal processing has not been performed, the white notch removal flow of the opposite color (S1603)
Transition to. The same applies to the removal of the white notch.
If the white notch removal processing is performed in step 1603), the white notch removal flow (S1603) is performed again. If the notch removal processing has not been performed, the flow shifts to the opposite color notch removal flow (S1602).

In the second example of the high-speed notch removal method using the run length presented so far, that is, the notch removal method for verifying only three lines, basically only three lines are verified. In addition, by combining the notch patterns, high-speed image compression processing can be realized several tens times that of the first method, ie, several hundred times that of the conventional image compression / high-quality conversion processing using pixel-by-pixel pattern matching. Things become possible.

In this method, the code length is long and the compression ratio is reduced.
Since the effect of removing the code (horizontal code) is high, it is possible to realize the compression code conversion without significantly inferior in the compression ratio to the first method.

Further, the present method is based on software processing using a CPU, and is performed by a conventional general PC, Uni, or the like.
The present invention can be easily applied to the compression processing from pixels of an x-machine or the like or a fax or an image storage device to an MR or MMR code only by replacing software.

Here, the method of compressing an image from pixel data to compressed data such as MR and MMR in the present invention will be described again together with its function and operation. First, image data is input. The image data is input by a method of inputting using a scanner, a keyboard, a mouse, or the like, or a method of inputting image data from an external information device or the like. Next, the pixel data of the input image is converted into run-length data. Next, an isolated point pattern or a notch pattern is extracted from the run-length data, and the pattern is removed. Next, MR and MMR are performed on the run-length data having been subjected to the notch pattern removal processing.
A conversion process to compressed data such as codes is performed. After the completion of the conversion process, the compressed data is transferred to an external information device or stored in an image storage device.

In the case of a system having an image decompression processing method,
The compressed data such as MR and MMR codes are restored to pixel data. After the completion of the restoration process, the pixel image data is output to a display, a printer, or the like. Alternatively, it is stored in an image storage unit or output to an external information device.

Two methods are proposed for image compression from the pixel data to compressed data such as MR and MMR. The first method is a method in which a notch pattern to be removed from an original image is specified in advance, and a notch removal process is performed on the notch pattern faithfully with respect to run-length data. In the above method, notch removal and the accompanying improvement in image quality have a great effect, but there is a problem that the processing speed is reduced as the number of notch patterns increases. However, even in the above method, speedup of several times to several tens times can be expected compared to the conventional method of directly performing pattern matching on pixel data.

Further, in the second method, the number of run-length data lines for performing pattern matching is limited to three lines, thereby realizing speedup several times to several tens times that of the first method. It realizes several hundred times faster processing.

Thus, according to the present invention, in image compression processing for converting pixel data of an image into compressed data such as MR, MMR, etc., the same quality and compression ratio as those of the methods used so far are obtained. While realizing, it is possible to realize a high-speed image compression process several hundred times higher than the conventional method.

The present invention relates to claims 1, 2 and 3.
However, in addition to these, the following items also constitute embodiments of the present invention.

(1) In the image compression apparatus according to the first aspect, as a pattern to be detected, a horizontal isolated point, a notch,
An image compression apparatus characterized by including detection of cornering or missing corner patterns and their removal.

(2) In the image compression apparatus according to the first aspect, as the pattern to be detected, a vertical isolated point, a notch,
An image compression apparatus characterized by including detection of cornering or missing corner patterns and their removal.

(3) In the image compression apparatus according to the first, second or third aspect, as a condition for determining an isolated point and a notch pixel, an isolated point whose approximate pixel length is equal to or less than n pixels (n is a natural number of 1 or more) is used. Alternatively, an image compression apparatus characterized in that only notches are to be removed.

(4) The image compression apparatus according to claim 1 or 2, wherein only an original image is targeted as an image to be subjected to pattern detection and removal / correction.

(5) The image compression apparatus according to claim 1 or 2, wherein the image to be subjected to pattern detection and removal / correction is repeatedly performed on an original image and a converted image obtained by removing / correcting the original image. Image compression device.

(6) The image compression apparatus according to claim 1 or 2, wherein only a pattern in one of black pixels and white pixels is to be removed.

[0093]

As described above, according to the present invention, in image compression processing for converting pixel data of an image into compressed data such as MR, MMR, etc., the quality is almost the same as that of the techniques used so far. It is possible to realize a high-speed image compression process several hundred times higher than the conventional method while realizing a high compression ratio.

[Brief description of the drawings]

FIG. 1 is a configuration diagram of an image compression processing system according to an embodiment of the present invention.

FIG. 2 is a PAD diagram showing a flow of an image compression process according to an embodiment of the present invention.

FIG. 3 is a PAD diagram showing a flow of an image restoration process according to an embodiment of the present invention.

FIG. 4 is an example of run length data.

FIG. 5 is an explanatory diagram of a first method in a high-speed notch removal method using run length.

FIG. 6 is a pattern diagram of a second method in a high-speed notch removal method using run length.

FIG. 7 is an explanatory diagram showing a black upward notch determination in a second method.

FIG. 8 is an explanatory diagram showing a white upward notch determination in the second method.

FIG. 9 is an explanatory diagram showing a black downward notch determination in the second method.

FIG. 10 is an explanatory diagram showing a notch removal method for run-length data in a second method.

FIG. 11 is an explanatory diagram illustrating a method of performing notch determination only on an original image.

FIG. 12 is a flowchart showing black notch removal in line units in the second method.

FIG. 13 is a flowchart showing white notch removal on a line-by-line basis in the second method.

FIG. 14 is an explanatory diagram showing a one-pass processing method of black / white notch removal in the second method.

FIG. 15 is a flowchart showing a one-pass processing method of black / white notch removal in the second method.

FIG. 16 is a pattern diagram 1 of a first method in a high-speed notch removal method using run length.

FIG. 17 is a pattern diagram 2 of a first method in a high-speed notch removal method using run length.

FIG. 18 is a pattern diagram 3 of a first method in a high-speed notch removal method using run length.

[Explanation of symbols]

101 image input unit 102 pixel data → run-length data conversion processing unit 103 run-length data → run-length data high compression data conversion processing unit 104 run-length data → compression data conversion processing unit 105 compressed data → pixel data restoration processing unit 106 image communication IF unit 107 Image storage unit 108 Image output unit

Claims (3)

[Claims] 1. An image input means for capturing image data,
An image processing apparatus comprising: run-length data conversion means for converting an image into run-length data expressed by a pixel color and its dot length for each line; and image compression means for converting the run-length data into a compressed image. At least one of an isolated point pattern, a notch pattern, a cornering or a missing corner pattern is obtained from the run-length data output from the data conversion means.
Pattern detecting means for detecting one pattern, and pattern removing / correcting means for removing or correcting the detected pattern from the run-length data by an output of the pattern detecting means. An image compression apparatus, wherein an output is input to the image compression means. 2. An image input means for inputting image data,
An image processing apparatus comprising: run-length data conversion means for converting an image into run-length data expressed by a pixel color and its dot length for each line; and image compression means for converting the run-length data into a compressed image. Pattern detection means for detecting an isolated point pattern and a notch pattern in the line direction with respect to run-length data output from the data conversion means by verifying three lines including a line before and after the corresponding line; An image compression apparatus comprising: a pattern removal unit that removes the detected pattern from the run-length data according to an output of the detection unit; and inputting an output from the pattern removal unit to the image compression unit. 3. An image input means for inputting image data,
In an image processing apparatus having run-length data conversion means for converting an image into run-length data expressed by black pixels and white pixels and its dot length for each line, and image compression means for converting run-length data to a compressed image, Pattern detection means for detecting an isolated point pattern and a notch pattern in the line direction with respect to run-length data output from the run-length data conversion means by verifying three lines including a line before and after the corresponding line; A pattern removing unit that removes the detected pattern from the run-length data according to an output of the pattern detecting unit. The pattern detecting unit and the pattern removing unit
The detection and removal of the isolated point pattern and the notch pattern in the black pixel and the white pixel are performed in parallel by the single verification of the three lines provided for the black pixel and the white pixel, respectively. An image compression device, wherein the output of the image compression means is input to the image compression means.