JPH05324815A - Compression system and output system for multivalued image - Google Patents

Abstract

PURPOSE:To provide a new system for the compression and restoration output of the multivalued image which can perform high-speed processing with both compressibility and reproducibility. CONSTITUTION:An input part 100 performs sampling and edge detection for thinning out pixels of many-valued image information and multivalued information at sampling points and binary information on the position and attribute of an edge are stored in memories 121, 122, and 123, respectively. An output part 140 inputs the pieces of stored information from the respective memories and interpolates the multivalued information on pixels between the sampling points, but performs linear interpolation at sample points where the edge is not present and distributes the multivalued information at sample points to both the sides between samples points where the edge is present. An edge having an attribute 'steep', however, is corrected by edge emphasis.

Description

Detailed Description of the Invention

[0001]

BACKGROUND OF THE INVENTION 1. Field of the Invention The present invention relates to a compression method and an output method for monochrome or color multi-valued image information.

[0002]

2. Description of the Related Art The following conventional techniques are known for compressing or outputting image information.

Prior art (a ) A method of compressing image information by redundancy compression coding, storing or transmitting it, and decoding it at the time of output.

Prior art b : A method of storing binary image information by thinning out pixels and interpolating the thinned pixels at the time of output.

Prior art c Only edge information such as characters is stored, an edge is drawn on a bit map memory based on this edge information, and black pixels are filled between the edges while scanning the bit map memory to obtain a binary value. System for reproducing images (Japanese Patent Laid-Open No. 53-410)
17, JP-A-3-22186).

Prior art d : A memory for separately storing edge information and gradation information is provided for each pixel, both memories are read at the same time, and the inside of the edge obtained by controlling the flip-flop with the edge information is a high-luminance pixel. An image display device that displays a binary image and a mixed image such as a figure with gradation by logically adding the embedded binary information and gradation information (Japanese Patent Laid-Open No. 3-296).
092).

[0007]

In an image transfer apparatus, a facsimile, a printer, an image database system, etc. that handles monochrome or color multivalued images, the capacity of the frame memory for storing multivalued image information becomes enormous. Is a problem. For example, when the color multi-valued image information of 300 DPI is directly stored in the frame memory, the information amount per pixel is 24 bits (R, G, B).
With each 8 bits), a huge memory capacity of 24 MB is required to store information for one A4 original.

Therefore, it is very important to compress the multi-valued image information by any method to reduce the data amount in order to reduce the memory capacity. Further, reduction of the data amount is important from the viewpoint of speeding up the transmission and output processing.

[0009] However, the above-mentioned conventional technique a can achieve both the data compression rate and the reproducibility of multi-valued image information, but the encoding process and the decoding process are complicated and high-speed processing is not easy. The processing system also becomes complicated and expensive. The prior art b may be applicable to multi-valued image information, but if the pixel thinning rate is increased, the edge of the original image and the reproducibility are extremely deteriorated, so that the quality of the restored image is deteriorated. It is difficult to achieve a significant reduction in data volume without doing so. Further, the conventional techniques c and d do not meet the purpose of data-compressing multi-valued image information and reproducing the multi-valued image information from the compressed information for output.

Therefore, it is an object of the present invention to provide a new compression method and output method for monochrome or color multi-valued image information, which is compatible with both compression rate and reproducibility and can be easily processed at high speed.

[0011]

According to the invention of claim 1, multivalued image information is sampled for edge detection and pixel thinning, and binary information indicating the position of the edge and sample points are sampled. A set of multi-valued information is generated as compressed information. According to the invention of claim 2, the same compression information is generated, but regarding the edge, in addition to the binary information indicating the position of the edge, information indicating whether or not the edge is steep is also generated.

According to the third or fourth aspect of the invention, color conversion is performed on color multi-valued image information, and pixel thinning is performed on the obtained brightness multi-valued information and color difference multi-valued information. 5. The sampling is performed and an edge is detected with respect to multivalued luminance information, and a set of binary information indicating the position of the edge and multivalued information of luminance and color difference at the sample point is generated as compressed information. According to the invention, the pixel thinning rate for the multi-valued color difference information is set to be larger than the pixel thinning rate for the multi-valued luminance information.

According to the invention of claim 5, compression information similar to that of the invention of claim 3 is generated, but regarding the edge, in addition to the binary information indicating the position of the edge, whether the edge is sharp or not is determined. The binary information shown is also generated.

According to the sixth aspect of the invention, the compression information according to the first aspect of the invention is used as input information, and when there is no edge between adjacent sample points, multi-valued information for pixels thinned out between adjacent sample points. Is linearly interpolated, and if there is an edge between adjacent sample points, the multi-valued information for the adjacent sample points is distributed by dividing the multi-valued information of the adjacent sample points on both sides of the edge. Interpolate,
By the above processing, multi-valued image information without pixel thinning is output.

According to the invention of claim 7, the compression information according to the invention of claim 2 is used as input information, and when there is no edge between adjacent sample points, multi-valued pixels for which thinned-out pixels between adjacent sample points are present. If information is linearly interpolated and there are edges between adjacent sample points, multi-valued information for adjacent sample points is distributed by dividing the multi-valued information of adjacent sample points on both sides of the edge. Information is interpolated, but when the edge is a steep edge, correction for edge enhancement is applied to the interpolated multi-valued information, and multi-valued image information without pixel thinning is output by the above processing.

According to the invention of claim 8, the compression information according to the invention of claim 3 or 4 is used as input information, and regarding color difference, multi-valued information for pixels thinned out between adjacent sample points is linearly interpolated, Regarding brightness, if there is no edge between adjacent sample points, linearly interpolate the multi-valued information for the decimated pixels between adjacent sample points, and if there is an edge between adjacent sample points, By interpolating the multi-valued information of the adjacent sample points on both sides of the boundary, the multi-valued information for the pixels thinned out between the adjacent sample points is interpolated, and the multi-valued information of the luminance and color difference obtained by the above processing is obtained. By performing reverse color conversion on the other hand, multi-valued color image information without pixel thinning is output.

According to the invention of claim 9, the invention of claim 5
The input information is the compression information by
Lines of multi-valued information for thinned pixels between tangent sample points
Shape interpolation, with regard to luminance, edge between adjacent sample points
If no image exists, the thinned image between adjacent sample points
Linearly interpolate multivalued information for primes and
If there is an edge, it will be next to it on either side of the edge.
By assigning multi-valued information of contact sample points,
Interpolate multi-valued information for thinned pixels between pull points
However, if the edge is steep,
Correction is applied to the interpolated multi-valued information and the above processing is performed.
Inverse color conversion for multivalued information of luminance and color difference obtained from
By applying the
Output image information.

[0018]

When the multivalued image information is compressed according to the first aspect of the invention, the multivalued information amount is reduced by the pixel thinning rate. The compressed information includes binary information of the edge position, but since this is 1 bit per pixel, the amount of information is small.
According to the invention of claim 2, the binary information about the edge is 1
It increases to 2 bits per pixel, but the amount of information is still small.

According to the invention of claim 6 or 7, multi-valued image information without pixel thinning is restored and output from the compressed information according to the invention of claim 1 or 7 by data interpolation. Since it is controlled based on the information, it is not possible to reproduce the fine gradation information of the original image without leaking it, but even if the pixel thinning rate during compression is chosen to be quite large, it is possible to restore multilevel image information of practically sufficient quality. it can. This means that it is possible to select a considerably large pixel thinning rate at the time of compression. Therefore, according to the invention of claim 1 or 2, the pixel thinning rate is set sufficiently large and the compression information is stored. It is possible to significantly reduce the memory capacity for and to ensure necessary and sufficient reproducibility.

When the multivalued image is compressed according to the second aspect of the invention and the multivalued image is reproduced and output according to the seventh aspect of the invention, a sharp edge on the original image is emphasized on the reproduced image. Therefore, it is possible to reproduce an image with good visibility of edges such as characters.

According to the first or second aspect of the present invention, since the compression processing is a combination of simple sampling and edge detection, the processing means can be made very simple as compared with the redundancy compression coding method. In addition, it becomes easy to speed up the process. Similarly, according to the invention of claim 6 or 7,
Since the image restoration output process is a simple data interpolation process, the processing means can be simplified and the process speed can be facilitated as compared with the redundancy compression encoding method.

For color multi-valued image information, the invention of claim 1 or 2 can be applied to the compression for each color component, and the invention of claim 6 or 7 can be applied to the decompression output. is there. However, according to the third, fourth, and fifth aspects of the invention, in the case of a color image, since the high-frequency component of the color difference information has a smaller effect on the image quality than the luminance system data, the compression ratio of It is even more advantageous.

That is, according to the third, fourth or fifth aspect of the invention, edge detection is performed only on the luminance information obtained by color-converting the color multi-valued image information, and the binary information is stored, so that each color component is stored. The amount of edge information can be reduced to one-third as compared with the case of storing edge information. Further, according to the invention of claim 4, by increasing the thinning rate for the color difference information as compared with the luminance information,
The reduction rate of the multivalued information amount can be increased as compared with the case where the luminance information and the color difference information are thinned out at the same ratio and stored. And
According to the invention of claim 8 or 9, a color multi-valued image of practically sufficient quality can be reproduced from such compression information.

Color compression or inverse color conversion processing is required at the time of compression or reproduction output according to the invention of claim 3, 4, 5, 8 or 9, and this processing can be performed at high speed by an extremely simple means. .. Therefore, the color multi-valued image can be compressed and reproduced and output at high speed with a simple processing system.

[0025]

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

Embodiment 1 FIG. 1 is a block diagram showing a device configuration according to this embodiment. 100 is an input unit, 120 is a frame memory unit, 14
Reference numeral 0 is an output unit, and 160 is a CPU. The input unit 100 includes a sub-sampling unit 101 and an edge detecting unit 102, and the frame memory unit 120 includes a multi-value information byte map memory 121, an edge position information bitmap memory 122, and an edge attribute information bitmap memory 123. The output unit 140 includes a decoder unit 141 and a data interpolation unit 1.
It consists of 42. The CPU 160 controls each unit and, if necessary, each memory 12 of the frame memory unit 120.
Information is written to 1 to 123.

Monochrome multi-valued image information (or multi-valued image information for one color of a color multi-valued image) is externally input to the input unit 100 as 8-bit / pixel digital data. For this multi-valued image information, the sub-sampling unit 101
Performs sampling for pixel thinning, and at the same time, the edge detection unit 102 performs edge detection. Multi-valued information (8 bits / pixel) from the sub-sampling unit 101 to sample points
Is output, which is written in the byte map memory 121 having an 8-bit / pixel structure.

The edge detection unit 102 outputs 1-bit edge position information and 1-bit edge attribute information corresponding to one pixel, and these are written in the bit map memories 122 and 123 of 1-bit / pixel configuration, respectively. When the edge position information is "1", it means that the pixel is an edge, and when it is "0", it means that it is not an edge. When the edge attribute information is "1", it means a sharp edge with a large gradation change, and when it is "0", it means an ordinary edge with a small gradation change.

Sampling (pixel thinning) by the sub-sampling unit 101 will be described with reference to FIG. (A)
Is a pixel array before sampling (pixel array of the input image). (B) shows a state of sampling (referred to as 1/4 sub-sampling) when pixel thinning is performed so as to reduce the number of pixels to 1/4, and one shaded pixel in 2 × 2 pixels is shown. Are sampled and the remaining 3 pixels are decimated. (C) shows a state of sampling (referred to as 1/16 sub-sample) when pixel thinning is performed so as to reduce the number of pixels to 1/16, and one shaded pixel in 4 × 4 pixels Only sampled and the remaining 1
Five pixels are thinned out.

FIG. 3 is a block diagram of the edge detector 102. Input data from the outside and input data immediately before it are latched in the latch circuits 171, 172, and the absolute value of the gradation difference is obtained by the subtractor 173. Comparator 174
Thus, the absolute value of this gradation difference is compared with the threshold values th1 and th2 (where th1 <th2) set in the latch circuit 176 by the CPU 160. One output s1 of the comparator 175 is edge position information, and the other output s2 is edge attribute information. s1 is “1” when the gradation difference absolute value ≧ th1 (when it is an edge), and is otherwise “1”.
0 ”. S2 is“ 1 ”when the gradation difference absolute value ≧ th2 (when it is a steep edge), and“ 1 ”when it is not.
It becomes 0 ".

Although the one-dimensional edge detection focusing only on the main scanning direction of the image is performed here, the two-dimensional edge detection focusing on the sub-scanning direction may be performed.

The output section 140 is a frame memory section 120.
Information stored in each of the memories 121, 122, 123 is read out and multi-valued image information without pixel thinning is reproduced to obtain 8 bits /
It is output to the outside as pixel data. The decoder unit 141 of the output unit 140 interprets the input information from the bit map memories 122 and 123 according to the following Table 1, and the data interpolation unit 1
Reference numeral 42 interpolates multivalued information for the images thinned out by a method according to the interpretation result.

[0033]

[Table 1]

When there is no edge between the adjacent sample points, the multivalued information for the thinned pixels is interpolated by linear interpolation from the multivalued information of the adjacent sample points and the distance between them. FIG. 4 is an explanatory diagram of this case. (A) and (c)
Indicates multi-valued information (gradation) of adjacent sample points read from the byte map memory 121. (B) shows output multi-valued information with respect to the input multi-valued information shown in (a), and (d)
Indicates output multilevel information corresponding to the input multilevel information shown in (c). In (b) and (d), the broken line is linearly interpolated according to the multivalued information of the adjacent sample points and the distance between them.
The multi-valued information for the thinned pixels is shown.

If there is an edge between adjacent sample points, the multi-valued information of the adjacent sample points is distributed to both sides of the edge as a boundary to interpolate the multi-valued information for the thinned pixels. However, if the edge is a steep edge, the interpolation data is corrected to emphasize the gradation change of the edge.

FIG. 5 is an explanatory diagram of data interpolation when there is a normal edge that is not sharp between adjacent sample points.
(A) and (c) show adjacent multi-valued information (gradation) input from the byte map memory 121. (B) shows the output multilevel information for the input multilevel information shown in (a), and (d) shows the output multilevel information for the input multilevel information shown in (c). In (b) and (d), the broken line indicates the interpolated multivalued information. As understood from this example, the multi-valued information of the front sample point is used as the interpolation data for the pixel before the edge, and the multi-valued information of the rear sample point is used as the interpolation data for the pixel after the edge.

FIG. 6 is an explanatory diagram of data interpolation when there is a sharp edge between adjacent sample points. (A) and (c)
Indicates input multi-valued information, and (b) and (d) indicate output multi-valued information corresponding thereto. The broken lines in (b) and (d) are interpolated multi-valued information. As understood from the example shown in (b), the interpolated multi-valued information is corrected so as to emphasize the gradation rise at the edge position in the section where the gradation rises. On the other hand, as can be seen from the example shown in (d), the interpolated multi-valued information is corrected so as to emphasize the gradation lowering at the edge position in the gradation lowering section.

FIG. 7 shows a state of image input → compression → reproduction. (A) is input multi-valued image information, (b) is 1/16
Multivalued information after sub-sampling, (c) edge position information,
(D) is reproduction multi-valued information. As can be understood by comparing (a) and (d), since the edge information is extracted and used in data interpolation, it is not possible to completely reproduce even small variations in the original image. It is possible to reproduce a multivalued image almost equivalent to.

Here, the memory capacity of the frame memory unit 120 will be described. In the case of 1/16 sub-sampling, the number of pixels of multi-valued information is reduced to 1/16 and the edge information is 2 bits per pixel, so the required memory capacity per pixel of the input image is 8 bits x 1 / 16 + 2 bits = 2.5 bits. Therefore, an A4 size image is 300 dpi
When inputting with, the memory capacity of 8 MB is required when the compression is not performed, but the memory capacity can be reduced to 2.5 MB. It is possible not to extract the edge attribute information, but in this case, the required memory capacity is further reduced.

As shown in Table 1, when the edge position information is "0" and only the edge attribute information is "1",
The decoding unit 141 stores the contents of the bitmap 123 in the CPU.
It is interpreted as a bitmap drawn by 160. In this case, the data interpolating unit 142 regards the bit “1” of the bitmap memory 123 as an edge, and performs a paint process, etc., in which a space between the odd-numbered edge and the even-numbered edge is filled with black pixels. The image is output as multi-valued information.

Embodiment 2 In this embodiment, a color multi-valued image is compressed and stored and reproduced, and its apparatus configuration is shown in FIG. 200 is an input unit, 2
20 is a frame memory unit, 240 is an output unit, and 260 is C
It is PU. The input unit 200 includes a color conversion unit 201, a sub-sampling unit 202, and an edge detection unit 203. The frame memory unit 220 is a byte map memory 2 for multilevel information.
1, 222, 223, edge position information bitmap memory 224, and edge attribute information bitmap memory 2
It consists of 25. The output unit 140 includes a decoder unit 241, a data interpolation unit 242, and a reverse color conversion unit 243. CPU
The control unit 260 controls each unit and, if necessary, writes information to the memories 221 to 225 of the frame memory unit 220.

The input unit 100 is provided with digital data (R, G,
8 bits for each B). With respect to this multivalued image information, the color conversion unit 201 performs color conversion from the RGB system to the luminance / color difference system, and the color difference (Cr = RY, Cb =
The BY information and the luminance (Y) information are each generated as 8-bit digital data. Sub-sample part 2
Reference numeral 02 performs sampling for pixel thinning on the color difference information and the luminance information, and outputs information on the sample points. This is written in the corresponding byte map memories 221, 222, 223 of 8 bits / pixel structure.

The edge detection unit 203 performs edge detection on the luminance information and outputs 1-bit edge position information and 1-bit edge attribute information for each pixel. This edge information is a bit map memory 22 having a 1-bit / pixel configuration.
4 and 225, respectively. Even if the color difference component cannot reproduce a high frequency, the influence on the reproduced image quality is smaller than that of the luminance component. Therefore, even if only the edge information of the luminance component is stored, there is no practical problem.

The required memory capacity will be described below.
When the sampling rate is selected so that the number of pixels is reduced to 1/16 for each of luminance and color difference, the required memory capacity per input pixel is 8 bits x 1/16 x 3 + 2 bits = 3.5 bits. Become. That is, in the case of A4 color image of 300 dpi,
Without compression, a memory capacity of 24 MB is required, but 3.5 MB is sufficient.

When image reproduction with higher accuracy is required, only the sampling rate (pixel thinning rate) for the luminance component can be lowered. For example, for the luminance component, the sampling rate is set to 1 so that 4 pixels are thinned out to 1 pixel.
/ 4, and the sampling rate for the color difference component is 1 /
If it is set to 16, the memory capacity required for each pixel of the input image is 2 bits for the luminance component, so that it is increased to 5 bits as a whole. Even with this, the required memory capacity can be significantly reduced as compared with the case without compression.

Next, the output section 240 will be described. The output unit 240 includes the memories 221 of the frame memory unit 220.
~ 225 read the stored information, RG without pixel thinning
The B-system color multi-valued image information is reproduced and output to the outside as data of 24 bits / pixel (8 bits for each of R, G and B).

The decoder section 241 of the output section 240 receives the input information from the bit map memories 224 and 223 as shown in Table 1 above.
The data interpolation unit 242 switches the processing contents according to the method according to the interpretation result. The processing of the data interpolation unit 242 is as follows.

Regarding the color difference information, multi-valued information for pixels thinned out between adjacent sample points is always interpolated by linear interpolation.

Regarding the luminance information, when there is no edge between adjacent sample points, multi-valued information for pixels thinned out by linear interpolation is interpolated, but when there is an edge,
By distributing the multi-valued information of the adjacent sample points to both sides thereof, the multi-valued information for the thinned pixels is interpolated. However, if the edge is a steep edge, edge enhancement correction as shown in FIG. 6 is performed.

It should be noted that only the edge position information may be extracted and stored as the edge information. In this case, the required memory capacity can be further reduced because the bitmap memory 225 becomes unnecessary.

The inverse color conversion unit 243 converts the multivalued information of luminance and color difference obtained as described above into RGB multicolored image information and outputs it to the outside.

When the edge position information is "0" and the edge attribute information is "1", the content of the bit map memory 225 is interpreted as a bit map drawn by the CPU 260. In this case, the data interpolation unit 242 considers the “1” bit of the bitmap memory 225 as an edge,
The multi-valued information of the binary image, which has been subjected to paint processing for filling the spaces between the odd-numbered edges and the even-numbered edges with black pixels, is output to the outside.

[0053]

As is apparent from the above description, according to the inventions of claims 1 to 9, the amount of information of a monochrome or color multi-valued image is significantly compressed by simple and high-speed processing, and it is necessary to store it. It is possible to drastically reduce the required memory capacity, and it is possible to reproduce a multi-valued image of practically sufficient quality by simple and high-speed processing. Especially claim 2
According to the inventions of 5, 7, and 9, the reproducibility of the edges of characters and the like included in the original image can be improved. According to the invention of claim 3, 4 or 5, the color multi-valued image information can be efficiently compressed by utilizing the properties of the color difference information and the luminance information.

[Brief description of drawings]

FIG. 1 is a block diagram showing a device configuration of a first embodiment.

FIG. 2 is a pixel array diagram for explaining sampling for thinning out pixels.

FIG. 3 is a block diagram showing a configuration of an edge detection unit.

FIG. 4A shows multi-valued information of adjacent sample points having no edge between them. (B) The data interpolation result is shown. (C) shows multi-valued information of adjacent sample points having no edge between them. (D) The data interpolation result is shown.

FIG. 5A shows multi-valued information of adjacent sample points having a normal edge between them. (B) The data interpolation result is shown. (C) shows multi-valued information of adjacent sample points having a normal edge between them. (D) The data interpolation result is shown.

FIG. 6A shows multivalued information of adjacent sample points having a steep edge between them. (B) The data interpolation result is shown. (C) shows multi-valued information of adjacent sample points having a steep edge between them. (D) The data interpolation result is shown.

FIG. 7A shows input multi-valued image information. (B) Multipoint information of sample points is shown. (C) Indicates edge position information. (D) Reproduction multilevel image information is shown.

FIG. 8 is a block diagram showing a device configuration of a second embodiment.

[Explanation of symbols]

 100 Input Unit 101 Sub-Sample Unit 102 Edge Detection Unit 120 Frame Memory Unit 121 Byte Map Memory for Multilevel Information 122 Bitmap Memory for Edge Position Information 123 Edge Attribute Information Bitmap Memory 140 Output Unit 141 Decoder Unit 142 Data Interpolation Unit 200 Input unit 201 Color conversion unit 202 Sub-sampling unit 203 Edge detection unit 220 Frame memory unit 221 Color difference (Cr) Multi-value information byte map memory 222 Color difference (Cb) Multi-value information byte map memory 223 Luminance (Y) Multi-value information Byte map memory 224 Edge position information bit map memory 225 Edge attribute information bit map memory 240 Output unit 241 Decoder unit 242 Data interpolation unit 243 Inverse color conversion unit

Claims (9)

[Claims] 1. A method for performing edge detection and sampling for pixel thinning on multi-valued image information, and generating a set of binary information indicating an edge position and multi-valued information of sample points as compressed information. Characteristic multi-valued image compression method. 2. The multi-valued image compression method according to claim 1, wherein binary information indicating whether the edge is steep is generated in addition to the binary information indicating the position of the edge with respect to the edge. 3. Multi-valued image information of color is subjected to color conversion, and multi-valued information of brightness and multi-valued information of color difference are subjected to sampling for pixel thinning and multi-valued information of brightness. A multivalued image compression method is characterized in that an edge is detected with respect to, and a set of binary information indicating an edge position and multivalued information of luminance and color difference of a sample point is generated as compression information. 4. The multi-valued image compression method according to claim 3, wherein a pixel thinning-out ratio for color-difference multi-valued information is made larger than a pixel thinning-out ratio for luminance multi-valued information. 5. The multi-valued image compression method according to claim 3, wherein binary information indicating whether the edge is steep is generated in addition to the binary information indicating the position of the edge with respect to the edge. 6. Multi-valued information for pixels decimated between adjacent sample points when there is no edge between adjacent sample points using the compressed information by the multi-valued image compression method according to claim 1 as input information. If there is an edge between adjacent sample points, the multi-valued information for the adjacent sample points is distributed to both sides of the edge, and the A multi-valued image output method, which interpolates information and outputs multi-valued image information without pixel thinning by the above processing. 7. Multi-valued information for pixels decimated between adjacent sample points when there is no edge between adjacent sample points using the compressed information by the multi-valued image compression method according to claim 2 as input information. If there is an edge between adjacent sample points, the multi-valued information for the adjacent sample points is distributed to both sides of the edge, and the Information is interpolated, but when the edge is a steep edge, correction for edge enhancement is applied to the interpolated multi-valued information, and multi-valued image information without pixel thinning is output by the above processing. Multi-valued image output method. 8. Using the compression information according to the multivalued image compression method according to claim 3 or 4 as input information, with respect to color difference, multivalued information for pixels decimated between adjacent sample points is linearly interpolated to obtain luminance. If there is no edge between adjacent sample points, linearly interpolates the multi-valued information for the thinned pixels between adjacent sample points, and if there is an edge between adjacent sample points, separates the edges. By interpolating the multi-valued information of the adjacent sample points on both sides, the multi-valued information for the pixels thinned out between the adjacent sample points is interpolated, and the multi-valued information of the luminance and the color difference obtained by the above processing is A multi-valued image output method that outputs color multi-valued image information without pixel thinning by performing reverse color conversion. 9. The multi-valued image compression method according to claim 5 is used as input information for input information, and for color difference, linearly interpolated multi-valued information for pixels decimated between adjacent sample points, and for luminance, If there is no edge between adjacent sample points, the multi-valued information for the thinned pixels between adjacent sample points is linearly interpolated, and if there is an edge between adjacent sample points, that edge is used as the boundary. By allocating multi-valued information of adjacent sample points on both sides, multi-valued information for pixels thinned out between adjacent sample points is interpolated, but when the edge is a steep edge, correction for edge enhancement is interpolated. Output multi-valued image information in color without pixel thinning by applying multi-valued information and inverse color conversion to multi-valued information of luminance and color difference obtained by the above processing. A multi-valued image output method characterized by.