JPH0342968A - Color picture information encoding system - Google Patents

Abstract

PURPOSE:To encode multilevel color picture information with a good reproducibility by encoding a binarized block obtained by binarizing each picture element data with the threshold set based on picture element data in a block and encoding values related to a maximum value and a minimum value of picture element data in the block. CONSTITUTION:A converting part 302 reads out 4-line components of the signal stored in a 4-line buffer 301 with the 4X4 size to segment 4X4 blocks and converts RGB signals to a signal L*a*b* of small inter-signal correlations by the look up table system where the memory table of a conversion table for L*a*b* is accessed by RGB signals. A maximum value and a minimum value or an average value and the difference between them are outputted as L (lightness) 308 from an encoding circuit 304. RGB signals inputted from a color scanner or the like are converted to the signal L*a*b* of small inter-signal correlations with respect to each unit block having the prescribed size, and the color picture of each block is expressed with lightness, structure, and color information based on the signal L*a*b*.

Description

DETAILED DESCRIPTION OF THE INVENTION (Technical Field) The present invention relates to a color image information encoding method for encoding multivalued color image information captured by a color original reading device, a color TV camera, or the like.

[Prior art]

Conventionally, when transmitting and storing image information, it has been common to take efficiency into account and suppress redundancy through encoding. In such encoding, most of the target image information is binary white/black or color information.

However, in recent years, image information has become more multi-valued and higher definition has been achieved, and color information has also become more multi-valued.

Therefore, it is necessary to encode multivalued color image information as well, but until now, the conventional method for black and white has been used to encode R (red), G (green), and B (blue). RGB is applied to each pixel, or RGB is applied to each pixel.
Consideration has been given to quantizing using the color correlation between the three primary colors. The former method is naturally inefficient and may also cause color misregistration in some cases. Also, in the latter case, color shift is less likely to occur, but since the correlation between the three primary colors of RGB is too strong,
High efficiency could not be expected.

Therefore, R (red), G (green), B (blue)
The three primary colors are converted into a signal form that has weaker inter-signal correlation and can be separated into brightness information and color information.Furthermore, the signal is cut out into small blocks, and for each block, the brightness within the block, the brightness within the block, It has been proposed to encode the structural information regarding the etch etc. into an information array representing the color information within the block.

That is, the R, B, and G signals are first treated as L'' in the Cl21976 uniform color space as an example of a signal form with smaller inter-signal correlation.
a″b″b″ signal, and further information in the small block is sent to L(
The information is encoded into an information format consisting of three elements: brightness), S (structural information), and C (color information).

FIG. 17 shows RGB-L"at b* in the target image.
It shows the conversion and cutting out of a 4 pixel x 4 pixel square block. In FIG. 17(a), 201
is a manuscript, 202 is a block, and 4 blocks are placed in order from the corner of the manuscript.
Blocks are cut out in ×4 size. Also 203
is one of the blocks, and indicates that the block includes an etched portion.

FIG. 17(b) shows the three primary colors (R,
G, B), and the RGB three primary colors are R as shown in the figure.
Sex appears only in .

FIG. 17(e) shows the case where the RGB signals shown in FIG. 17(b) are converted into L″a″b″.

Here, a conversion formula for converting from RGB to L"a"b" signals is shown below.

(However, Xo, Yo - Zo are the values of the reference white light.) (Problem H that the invention is trying to solve) In this way, the color image is divided into blocks of a predetermined size for each of L'a"b". and then convert it into a signal consisting of
By encoding these using vector quantization, efficient encoding of color images can be achieved.

However, this vector quantization is irreversible encoding, and unless optimal operations are performed during encoding and decoding, good color image reproduction will not be achieved.

This problem involves encoding multi-tone color image data.

This becomes even more important when decoding.

[Means to solve the problem]

The present invention has been made in view of the above points, and it converts color image information into a signal that has a weaker inter-signal correlation and can be separated into brightness information and color information, and also blocks each of the converted signals into a small block. In a color image information encoding method that cuts out a block and encodes each block, each pixel data is binarized using a threshold set based on the pixel data in the block, and a binarized block is encoded. This provides a color image information encoding method that uses structural information of each block by encoding values related to the maximum and minimum values of pixel data.

〔Example〕

The present invention will be explained below based on preferred embodiments.

FIG. 1 shows an embodiment of a circuit configuration for achieving encoding to which the present invention is applied. Reference numeral 301 is a 4-line buffer that temporarily stores RGB signals sequentially input line by line from a color scanner or the like in order to cut them out into the aforementioned blocks. That is, by reading out four lines of signals once stored in the four-line buffer 301 in a size of 4×4, a 4×4 block is cut out. 302 is RGB→L
This is an L"a"b" conversion section that performs "a"b" changes, and performs conversion operations based on the conversion formula shown above.

This L"a"b" conversion is realized by a look-up table method in which the memory table in which the conversion table for L"a"b1 is written is accessed by the RGB signal. In this way, the RGB signal is L with small correlation
It is converted into "a" and "b" signals.

303 is the 17th
X I l rX in the block L' in Figure (C)
121...+X2□X22. . . , is a multi-valued L1 signal output in the order of X44. 304 is this multi-value L
``This is an encoding circuit that encodes signals.

This encoding circuit 304 outputs the maximum value and minimum value of each block, their average value, and the difference between them as L (lightness).
It is output as 308.

Further, the encoding circuit 304 encodes the structure of the block based on the L'' signal. That is, the structure information 310 is rounded into a predetermined pattern.

311 and 312 are L” a” and b” converting units 302, respectively.
This is an averaging circuit that takes the averages a m and b II of each block of am and bm, which are the outputs of , and is composed of an adder and a divider.

313 is a quantizer that collectively quantizes the block average values of a'' and b''. This forms color information 314 for each block.

315 is L (lightness) 308. which is obtained as explained above. S (Structure) 310. This is a multiplexer that combines C (color information) 314 into one code for each block. 316 is the output signal of the multiplexer 315, that is, the encoded code.

In this way, the RGB signals input from a color scanner etc. are converted into L'a*bma*bm signals with low inter-signal correlation for each unit block of a predetermined size, and the color of each block is determined based on this L'a'b' signal. An image is represented by brightness, structure, and color information.

When decoding the code encoded in this manner and reproducing a color image, each area delimited by the edge of each block based on the structural information is painted differently with a color represented by the brightness and color information. This allows the color original image to be reproduced well.

In this embodiment, the RGB signals are shown as L″a″b″, but they can also be expressed as L′a″υ1 or NTSC YIQ, PAL,
YUV etc. can also be used.

In addition, human input signals are not limited to RGB, but depending on the sensor, Y (
Inputs such as yellow), G (green), and C (cyan) are also possible.

Further, although a'' and b'' are represented by average values, they may be stored in more detail.

Next, the operation of the encoding circuit 304 shown in FIG. 1 will be explained in detail.

FIG. 2 is a diagram showing the basic configuration of the encoding circuit.

In this embodiment, compression is performed for each 4×4 pixel block, and each pixel has a 6-bit (0 to 63) gradation. First, 4×4 blocks of the compressed image, 0 units,
i=4. Maximum level M from among 16 pixels of j=1 to 4
The detector 2 detects AX and the minimum level MIN. In the example of FIG. 2, MIN-4 and MAX=58. The average value of these maximum and minimum values is calculated by the calculator 3, and this average value (=
Using S t ) as a threshold, the comparator 1 binarizes the value of each pixel of the original block to obtain a binary block BIJ.

This binary block B is encoded using a vector quantization method and compressed.

For example, if 1 is black and O is white, the binarized block BIJ can be regarded as a black and white pattern. There are 2'6 patterns. On the other hand, N patterns (N is an integer less than 2'6) are determined in advance as reproduction patterns, and codes are assigned to these patterns.This set of reproduction patterns is called a codebook.

The basic concept of the encoder 4 of this method is that there are certainly 216 possible black and white patterns, but not all of these patterns will occur when an actual image is processed. do not have. Depending on the resolution of the image reading system, there are patterns that do not occur at all or patterns that rarely occur. Further, even if the pattern is replaced with another pattern, there are patterns whose deterioration is not visually perceptible as a reproduced image. Furthermore, a group of similar patterns can be reduced to one pattern.

In this way, by limiting the number of reproduction patterns, image data can be compressed while maintaining the quality of the reproduced image. In other words, what and how many patterns to register in the codebook as playback patterns depends on the quality of the playback image, including the quality of the playback image.
This is a major factor in compression efficiency.

FIG. 3 shows the codebook in this embodiment, and there are 24 patterns in total. All of the 218 black and white patterns to be reproduced are assigned to one of these 24 patterns, and the code of that pattern becomes the code of the input cover pattern. This code assignment is generally done so that distortion between the input cover pattern and the playback pattern is minimized.

In this way, 216 patterns are compressed into N (=24) patterns, so the compression ratio is j! Og221
6/ fLog2 N-16/jlog2 N,
When N=24, it becomes about 3.5.

This encoding can be achieved using a storage element such as a ROM. As shown in FIG. 4, each pixel of the binarized block BIJ is made to correspond to an address in the ROM. In this case, since there are 16 address lines, the capacity only needs to be 64 Kbytes or more. Then, a reproduction pattern code is written in each address of the ROM. In this way, encoding of the binarized block can be realized almost in real time. In FIG. 4, the address of the binarization pattern is t t to t oooooooooooo.

=53248, and code O of the reproduction pattern is written to this address.

Such a translation table is a lookup table (
The encoder 4 in FIG. 2 is an LUT for encoding a binary pattern BIJ configured in a ROM, and outputs a code CR of a reproduction pattern corresponding to each input.

On the other hand, the maximum value MAX and minimum value MIN as gradation information are compressed using the same concept.

First, some combinations of maximum and minimum values are registered in the codebook. As in this embodiment, each pixel has 6-bit gradation, which is 2080 ways.

In this embodiment, 128 combinations are registered as shown in FIG. 5, and each combination is assigned a code. The idea behind this registration is to register combinations with a small difference between the maximum value and minimum value in detail, and conversely, to register combinations with a large difference in a coarse manner. This means that for blocks with small differences, that is, blocks with low contrast, the level error during playback will be visually noticeable as deterioration, but for blocks with large differences, that is, blocks with edges etc., the level difference will be to some extent. This is because if it can be reproduced, even if there is an error in the absolute level, it will not be visually noticeable.

This maximum value and minimum value can also be encoded using a look-up table constructed from a ROM or the like. As shown in FIG. 6, the maximum value and minimum value are made to correspond to addresses. Each is 6 bits, so the ROM capacity is 4Kbytes.
It is fine as long as it is in group TE or above. For each combination of maximum and minimum values, the combination that results in the smallest distortion is searched from among the combinations registered in the codebook, and the code is written to the corresponding address.

The gradation data encoder 5 in FIG.
This can be achieved with In the example of FIG. 2, the code CT of the grayscale signal
=26 is output.

In this way, encoding of the image signal can be realized.
*Two codes of CT are obtained. Transmission and storage are performed in this compressed state.

Now, decoding will be explained next.

Since the code CI of the block pattern is encoded as a black and white binary pattern, even if it is decoded as is, it can only be decoded as a binary pattern. Therefore, 2
Gradation is reproduced by processing the value pattern with a spatial filter such as a low-pass filter and then decoding it.

This decoding process will be explained with reference to FIG.

Binary pattern BIJ corresponding to code CR is 1°XIO
The binary pattern Ek+ is expanded to =1 to 10.1 = 1 to 10. This is the kth time that a 1×7 spatial filter is applied to the binary pattern. If the size of the filter is 1×5, expansion to an 8×8 binary pattern is sufficient;
In the case of x9, it is necessary to expand to a 12x12 binary pattern.

The following two methods can be considered for expanding this 4×4 binary pattern to a 10XIO binary pattern.

One method is to extend the same value as the surrounding pixels of a 4×4 block outward in a direction perpendicular to the sides.
In this method, the direction of the edges of the black-and-white pattern is considered and the pattern is expanded in that direction.

The example in Figure 7 shows toxt expanded according to the second method.
This is the binary pattern Ekl of o.

Add 1x7 Gaussian F i 1t to this
Apply ar. Each element of the filter is m = (4°25.
70, 100, 70, 25.4).

As shown in FIG. 7, this is subjected to a sum-of-products operation with Ekl in the direction perpendicular to the edge to obtain the H-head.

Next, a proportional calculation is performed so that the maximum value of this HIJ becomes 63 and the minimum value becomes 0, and a gradation pattern S normalized by 63 is obtained.
Get a mouth.

The same process is performed for all the binary patterns registered in the code pattern in this way, and the Sth normalized gradation pattern is obtained in 63.

When the code CR is input, the decoder 6 in FIG. 2 outputs the gradation pattern SIJ instead of the corresponding binary pattern B.

This decoder 6 has a pull-up table (
By configuring a LUT for each pixel, it can be realized using a storage element such as a ROM.

On the other hand, the 10 gradations are decoded by a decoder 7 configured as an LUT that outputs the maximum value MAX and minimum value MIN of the codebook as they are.

Using each pixel of the gradation pattern normalized in this way and the decoded maximum value MAX and minimum value MIN, the calculation unit 8 performs the calculation R1j=StjN63/MAX+MIN to obtain the final decoded pattern RIJ. obtain.

In this embodiment described above, the maximum level and minimum level within a block are detected as gradation information, and compression is performed as a combination of the maximum value and minimum value, but as shown in FIG. After detecting the minimum value, the sum and difference are calculated by the calculator 9, and the sum and difference values are compressed as gradation information. This is based on the same concept as the embodiment shown in FIG. 2, and the encoder 10 configured with an LUT encodes the sum value and the difference sum as a combination.

On the other hand, during encoding, the decoder 11 is configured to output the maximum value and minimum value from the code instead of the sum value and difference value.
By using an up table, subsequent processing can be performed in the same manner as in the embodiment shown in FIG. Here, the sum value is the sum of the minimum value and the maximum value, but by doing so, this value can be directly used as a threshold for binarization of 0 original image signals.

In this way, rather than compressing directly as a combination of maximum and minimum values, it is possible to compress more efficiently by compressing as a combination of differences and sums. This is 4×4 when taking the difference.
within the block, the difference is 63, that is, the maximum value is 6
3. The case where the minimum value is 0, which is a large value, rarely occurs, so the range of the difference value becomes small. This is similar to the idea of predictive hakama.

This is particularly effective in systems that handle low-contrast images such as photographs as input images, or in systems where the MTF of the reading device is small.

As explained above, a gradation image is divided into blocks, binarized using an appropriate threshold value, and then encoded. During decoding,
A spatial film such as a Gaussian Filter is decoded as a normalized gradation pattern obtained by applying a binary pattern, and each pixel of the normalized gradation pattern is simultaneously encoded and decoded according to the maximum and minimum values in the decoded block. By performing proportional calculations on , it has become possible to achieve efficient data compression for reproducing gradation images at low cost.

In order to reproduce a complete image by decoding multi-tone image data encoded as described above, it is necessary to consider data representing structure and data representing density. There are cases where you only want to understand the outline of an image, or cases where you want to quickly understand the outline. The configuration that makes this possible will be explained next.

In this embodiment shown in FIG. 9, the compression is 4
This is done for every ×4 pixel block, and each pixel is 6b
It has gradations of (0 to 63).

First, 4×4 blocks O+j, i=1 to 4 of the compressed image
．． Maximum level MAX from 16 pixels of j==+1 to 4
, the minimum level MIN is detected by the detector 2. In the example of FIG. 9, MIN-4 and MAX=58. The sum and difference calculator 21 calculates the average value (MAX+MIN)/2 and the difference value (M
AX-M I N) are calculated respectively. This average value (
Using ÷31) as a threshold, the original block OIJ is binarized by the comparator 1 to obtain a binary block BIJ. This binary block BIJ is encoded and compressed by the vector quantization method in the encoder 4 as in FIG. 2.

The encoder 4 in FIG. 9 is an LUT for encoding a binary pattern B, which is configured in a ROM.

The encoder 4 outputs a code CR of a reproduction pattern corresponding to each input.

This code CR and the average value and difference value of the maximum value MAX and minimum value MIN, which are gradation information, are stored in a memory or transmitted as encoded data.

Here, the average value (MAX+MIN)/2 of the gradation information
can be regarded as the average level of the block, so it can be used as is as an image signal for a low-resolution display device such as a CRT 23.

This is effective when searching for images, editing images, etc. In addition, since the difference value can be regarded as an amount that represents the contrast within a block, by using it to determine the presence or absence of edges, it is possible to encode regional separation between line drawings such as characters and continuous tone drawings such as photographs. It can be done in the state.

Now, decoding will be explained next.

As with the example shown in the second diagram, the block pattern code CR is encoded as a black and white binary pattern, so even if it is decoded as is, it can only be decoded as a binary pattern. Therefore, in the same manner as described above, the gradation is reproduced by decoding the binary pattern after processing it with a spatial filter such as a low-pass filter using the method shown in FIG.

This decoder 6 can be realized with a storage element such as a ROM by configuring a look-up table (LUT) for each pixel that takes the code as an input address and outputs decoded pixel data, and can be implemented using a manual code. A gradation pattern SIj corresponding to the gradation pattern SIj is output.

On the other hand, the gradation information stored and transmitted as the sum and difference between the maximum value and the minimum value is processed by the sum-difference calculator 22, which performs the opposite calculation to the sum-difference calculator 21. Restored to MIN.

From each pixel of the gradation pattern normalized in this way and the restored maximum value MAX and minimum value MIN, RI J = SI J H63/ MAX +
The computation of M r N is performed by the computing unit 8 to obtain the final decoding pattern RIJ.

As explained above, a gradation image is divided into blocks, binarized with an appropriate threshold value, and then encoded. During decoding, G
Decoding as a normalized gradation pattern obtained by applying a spatial filter such as the Aussian Filter to a binary pattern, and performing proportional calculation for each pixel center of the normalized gradation pattern according to the maximum and minimum values in the block. The present invention, which reproduces a gradation image using a gradation image, has made it possible to perform efficient data compression at a low cost.

Also, as gradation data, the average value of the maximum and minimum values, which can be considered to represent the average level within the block, and the difference between the maximum and minimum values, which can be considered to represent the contrast within the block, are transmitted or stored in memory. Therefore, on a low-resolution display device such as a CRT, the coded state can be displayed using only the bits corresponding to the average value, which is effective for image searches and the like. In addition, the difference information can be used as the contrast within a block, that is, the amount to judge the presence or absence of edges.
Region separation of photographs, etc. is possible in the encoded state.

In the embodiments described above, -regular encoding is performed regardless of the content (density, contrast, etc.) of the original image. However, if the encoding method is adaptively changed depending on the image content company, encoding efficiency and reproducibility can be improved. An example will be explained below.

In FIG. 10, the compression is 4×
This is done every 4 pixel blocks, and each pixel is 6bi
It has gradations of t (0 to 63).

First, 4x4 blocks O+j, iyo 1-4 of the compressed image
．． j- The maximum - from 16 pixels from 1 to 4.

A detector 32 detects the level MAX and the minimum level MIN. In the example of FIG. 10, MIN=4 and MAX=58. This is calculated by the sum-difference calculator 33 to calculate the average value (MAX+MI
N)/2 and the difference value (MAX-M I N) are respectively calculated. Using this average value (=31) as a threshold, the original block 0muj is binarized by the comparator 31, and the binary block 81j is
get.

This binary block B is encoded by the encoder 34 using the vector quantization method and compressed as described above.

FIG. 11 shows the codebook for the example shown in FIG. 10, and there are 24 patterns in total. All of the 2'6 black and white patterns that can be reproduced are assigned to one of the 24 patterns, and the code of that pattern becomes the code of the input cover pattern. Generally, the codes are assigned so that the distortion between the human cover pattern and the playback pattern is minimized.

In this way, 216 patterns are compressed into N (316) patterns, so the compression ratio is 11, og216
/JZog2N=16/ILog2 N, which is N-1
6 becomes 4.

This encoding can be achieved using a storage element such as a ROM.

The encoder 34 in FIG. 10 is an LUT for encoding a binary pattern BIJ configured in a ROM, and outputs a code CRI of a reproduction pattern corresponding to each input.

On the other hand, the maximum value MAX and minimum value MIN as gradation information are compressed using the same concept.

First, some combinations of average values and difference values of maximum and minimum values are registered in the codebook. Similar to the present embodiment described above, when each pixel has 6-bit gradation, in this embodiment of the maximum value and minimum value, 1213 combinations are registered as shown in FIG. 12, and each combination is Code the code. The idea behind this registration is to register combinations with a small difference between the maximum and minimum values in detail, and conversely, to register combinations with a large difference in a coarse manner. This means that for blocks with small differences, that is, blocks with low contrast, the level error during playback is visually noticeable as deterioration, but for blocks with large differences, that is, blocks with edges, etc., the level difference is to some extent. This is because if it can be reproduced, even if there is an error in the absolute level, it will not be visually noticeable.

The encoding of the maximum value and minimum value can also be realized using a look-up table constructed from a ROM or the like. As shown in FIG. 13, the average value and the difference value of the maximum value and minimum value are made to correspond to the address. Since each bit is 6 bits, the capacity of the ROM only needs to be 4 Kbytes or more. For each combination of the average value and difference value of the maximum value and minimum value, search for the combination that results in the smallest distortion from among the combinations registered in the codebook, and write that code to the corresponding address. The encoder 35 for gradation data shown in FIG. 10 can be realized by such an LUT. In the example shown in FIG. 10, the tone signal code CT=92 is output.

In this way, encoding of the image signal can be realized, and CRt
: A total of 11 bits of code including 4 bit%Ct+near bit is obtained.

Incidentally, in the above explanation, a block of a gradation image is separated into two states: a binary block, and the maximum value and minimum value within the block, and each state is encoded. The former can be considered to be related to the resolution of image edges, etc., and the latter to be related to the gradation of the image.

It goes without saying that when this information is encoded, the longer the code length, the more faithfully it can be reproduced.
When the overall code length is determined, its distribution affects the quality of the reproduced image.

That is, by increasing the code lengths of the maximum and minimum values, gradation can be better reproduced, and by increasing the code length of the binarized block, edges can be better reproduced. For example, for images such as photographs, it is better to perform encoding that emphasizes gradation rather than edges, and for images such as text, it is better to emphasize reproduction of edges rather than gradation.

Therefore, we decided to use the aforementioned binarized block encoder 34 and maximum value and minimum value encoder 35 for photographic images.
Binarization block encoder 36 for character images in a similar manner
and a maximum value/minimum value encoder 37 is created.

The codebook of the binarized block encoder 36 is shown in FIG. 14, and compared to the codebook for photographic images shown in FIG. 11, the number of registered playback patterns is doubled to 32 patterns. That is, the code length increases by 1 bit to become 5 bits. On the other hand, the maximum value and minimum value are similarly encoded as a combination of a difference value and an average value using the code book shown in FIG. In this case, the number of registrations is 64, which is half that of the code book for photographic images shown in FIG. 12, and the number of codes is also reduced by 1 bit to abtt.

In this way, the binarized block is converted to a 5-bit code CR2 by the encoder 36, and the average value and the difference value of the maximum value and minimum value are converted to the encoder 36.
7 and is encoded into a 6-bit code Cta, resulting in a total code length of 11 bits, similar to that for photographic images.

Next, one of the codes CRI and CRj of the binarized block is selected by the selector 38 depending on the image tone of the block.

In other words, if the block is determined to be part of a photographic image, C
If RI is determined to be part of a character image, CR2 is output. Similarly, selector 3 also applies to the signs of the maximum and minimum values.
In step 9, CT2 is selected for a photographic image and CT2 is selected for a character image.

By the way, in this embodiment, the determination as to whether a block is a photographic image or a character image is made based on the difference value between the maximum value and the minimum value. In other words, the difference between the maximum and minimum values is considered to be small in low-contrast images such as photographic images, and conversely, the difference between the maximum and minimum values in blocks that include edges, such as text images, is considered to be large. I can do it. Then, the comparator 40 compares the difference value between the maximum value and the minimum value with a predetermined threshold value, and outputs this comparison to the respective selectors 38 and 39.
This is the select signal.

As shown in FIG. 16, the codes CR and CT obtained in this way are output with two bit allocations depending on the photograph and the text, but since the overall bit length of both is 1 lbit, the memory device 41 &: There is no need to change each code for storage or transmission, and the same configuration can be used.

Now, decoding will be explained next.

Since the code CR of the block pattern is encoded as a black and white binary pattern, even if it is decoded as it is, it can only be decoded as a binary pattern.Therefore, the method shown in FIG. The gradation is reproduced by decoding the image after processing it with a spatial filter such as a low-pass filter.

This decoder 42 can be realized with a storage element such as a ROM by configuring a look-up table (LUT) for each pixel, which uses the code as a manual address and outputs decoded pixel data, and reads the input code. The gradation pattern S corresponding to the gradation pattern S is output.

On the other hand, the gradation code CT is the maximum value M'AX of the codebook.
and the minimum value MIN are decoded by a decoder 43 configured as an LUT that outputs the minimum value MIN as is.

From each pixel of the gradation pattern normalized in this way and the decoded maximum value MAX and minimum value MIN, the calculation unit 14 performs the calculation RIJ-SIJN63/MAX+MIN to obtain the final value. Obtain a decoding pattern R.

As explained above, a gradation image is divided into blocks, binarized using an appropriate threshold, and then encoded and decoded using Ga
It is decoded as a normalized gradation pattern obtained by applying a spatial filter such as USSIAN F i 1 ter to a binarized pattern, and then normalized according to the maximum and minimum values in the simultaneously encoded and decoded block. Switching the code length of the binarized block and the distribution of the maximum and minimum value code lengths depending on the tone of the block when proportional calculation is performed on each pixel of the gradation pattern to restore the gradation image. in,
A highly efficient compressor that can handle all types of images can be realized at low cost.

In addition, making the overall bit length equal for each image level eliminates the need to change the configuration of the memory device for storage or the transmission device for data communication for each image level, and can be realized with a simple configuration and at low cost. This is effective for system configurations.

〔Effect of the invention〕

As explained above, according to the present invention, color image information is converted into a signal that has weaker inter-signal correlation and can be separated into brightness information and color information, and each of the changed signals is cut out into small blocks. In a color image information encoding method that encodes each pixel data separately, a binarized block is encoded by binarizing each pixel data with a threshold set based on the pixel data in the block. By encoding values related to the maximum and minimum values of data, the structure information of each block is obtained, so that multivalued color image information can be encoded with good reproducibility.

[Brief explanation of drawings]

FIG. 1 is a block diagram of the entire encoding section to which the present invention is applied, and FIG.
8, 9 and 10 are block diagrams of the encoding circuit, and FIGS. 3, 11 and 14 are diagrams showing examples of reproduction patterns, and FIGS. 4 and 5. 6, 12, 13 and 15 are diagrams showing the encoding operation, FIG. 7 is a diagram showing the decoding operation, FIG. 16 is a diagram showing the form of the encoded code, and FIG. 17 is a diagram showing the encoding operation. It is a diagram showing a color image conversion operation, in which 304 is an encoding circuit, 1 is a comparator, 2 is a detector, and 4 is an encoder. Code: O Code code = 2 Code: 3 Address al182183+BarB+zBz2Ba2B<z8
+5BuB3sB4sB14824834E144 data Figure 5 Figure 6 Figure 8 Hood: 0 Code: 1 Code = 2 Hood 3 Code: 4 Code = 5 Code: 6 Code = 7 Code 28 Code: 9 Code = 10 Hood: 11 Code = 12 Code 3 Code 2 14 Code: 15 Praise cannot exist Code = 4 Code 6 Code, 7 Code F9 Code: 10 Code, 11 Code, 13 Code F: 15 Code: 19 Code: 22 Code: 23 Code F = 24 Code = 25 1 code Kaya/Roguchi

Claims (3)

[Claims] (1) Converting color image information into a signal with weaker inter-signal correlation and separable into brightness information and color information, and
Cut each of the converted signals into small blocks,
In a color image information encoding method that encodes each block, each pixel data is binarized using a threshold set based on the pixel data in the block, and a binarized block is encoded. A color image information encoding method characterized in that structural information of each block is obtained by encoding values related to maximum and minimum values of . (2) The color image information encoding method according to claim 1, wherein the average value and difference between the maximum value and minimum value of pixel data within a block are encoded. (3) A color image information encoding method characterized by adaptively changing the code data length obtained by coding a binarized block and the code data length obtained by coding the maximum value and minimum value according to the image content.