CA1333421C - Method and apparatus for processing display color signal - Google Patents

Abstract

A color video signal processing method and apparatus wherein compressed display data of each pixel is formed of color identification code and color satura-tion data relative to the color or luminance of one pixel having the maximum saturation change in comparison with an adjacent pixel, and the original color data is reproduced from such display data. Furthermore, a plurality of representative colors are prepared per value of specific color, and the optimal color is selected per pixel to provide display data. A distri-bution of color appearance frequencies is employed for determining the representative colors, and such colors are so selected as to have the maximum color distances from one another.

Description

BACKGROUND OF THE INVENTION:
Field of the Invention:
The present invention relates to a method and an apparatus for processing color video signal.

Description of the Prior Art:
It has been known heretofore that, in computer graphics for example, a sufficient capability of display-ing 256 (= 28) colors per pixel is generally required.
In other words, the requisite is satisfied by allocating 8 bits to each pixel.
Recently, however, there exists another require-ment of processing or displaying a natural picture such as an image picked up by a video camera. In such a case, the conventional 256-color display capability is not sufficient to meet the requisite.
In the computer graphic system equipped with a display function of 2048 pixels x 2048 pixels or so, if 8 bits are allocated to each of red, green and blue in one pixel for increasing the color display capability per pixel, it follows that a total of 24 bits are needed for each pixel, and therefore the entire bits required are 24 bits x 2048 pixels x 2048 pixels = 100,663,296 bits = 12,582,912 bytes ~ 12 megabytes Thus, the display memory is rendered excessive in size.
In addition, an extremely long time is necessary for processing the image, and merely a few sheets of images can be saved even by employing a hard disk apparatus, whereby satisfactory adaptability for practical use fails to be attained.
For solution of the above problems, there is known a method of extending the color display capability by means of a color look-up table while leaving 8 bits unchanged per pixel. An exemplary color look-up table is constituted of memories having 4 addresses each com-posed of 8 bits and designated by 2-bit red and blue address signals respectively, and also 16 addresses designated by 4-bit green address signal, where 8-bit color saturation data is stored in each address. And a total of such 24 addresses are designated by 8-bit color look-up table address signals stored in the addresses corresponding to the individual pixels in the display memory. Accordingly, although merely 256 addresses are selectable in the color look-up table, a total of 16,777,216 values can be taken since the output of the color look-up table is composed of 24 bits in all. Therefore, despite the limited simultaneous display of merely 256 colors in one frame, it is rendered possible to display a total of 16,777,216 colors by rewriting the contents of the color look-up table during a blanking interval or the like. Consequently, due to the use of such color look-up table, many hues and gradations (color saturations or values of color compo-nents) are representable even with a small capacity of the display memory to eventually achieve display of a colorful image.
However, even in the technique mentioned, the kinds of colors for a natural picture may become insuf-ficient to eventually bring about lack of naturalness in the luster and tint. For example, in displaying lustrous fruits, merely 2 bits are allocated to red and blue respectively and, due to 4 gradations, quantization is rendered conspicuous with color changes appearing like stripes.
Besides the above, since merely 8 bits/3 - 2.7 bits on average are allocated with respect to each of red, green and blue, there arises a problem of insuffi-ciency in the gradations with further disadvantages including reduction of the resolution and deterioration of the picture quality.

1333~21 OBJECTS AND SUMMARY OF THE INVENTION:
It is therefore an object of the present inven-tion to provide a novel color video processing method and apparatus capable of eliminating the above-described drawbacks observed in the prior art.
In particular, a principal object of the present invention resides in the provision of a color video processing apparatus adapted to increase the number of colors generable simultaneously.
According to one aspect of the invention, there is provided an apparatus for processing color display signals, comprising:
color data generating means for generating a plurality of color data signals, each consisting of a predetermined number of bits and representing a respective one of a plurality of color components of a corresponding pixel;
transmission data generating means for generating a series of transmission data signals based on said color data signals and each representing a respective pixel and consisting of no more than said predetermined number of bits, said transmission data generating means including a central processing unit programmed for comparing at least two of said color data . ~

1333~1 signals, each representing a respective color component of a first pixel, with corresponding color data signals of a second pixel adjacent to and prece~;ng said first pixel to produce respective differential values for each said comparison, said central processing unit being further programmed for selecting as the transmission data signal representing said first pixel, one of said at least two color data signals of said first pixel having a maximum differential value with respect to said corresponding color data signals of said second pixel, and for combining in each said transmission data signal identification data identifying the respective color component and code data corresponding to a value of the respective color component: -receiving means for receiving said transmission data signals for said pixels;
a plurality of memory means each for storing display data signals representing a respective one of said color components;
means responsive to said identification data in said selected one of the transmission data signals which is received for renewing said display data signals stored in a corresponding one of said memory means, the display data signals which are renewed corresponding with a color 1~33~21 identified by said identification data ~nd having a renewed value corresponding to said code data in the received transmission data signal: and means for supplying the display data signals stored in said memory means to a display means.

BRIEF DESCRIPTION OF THE DRAWINGS:
Fig. 1 is a block diagram of an exemplary color video signal processing apparatus embodying the present invention;
Fig. 2 is a flow chart of a first method to encode the color signal compressed data according to the .' "!}

1333~21 invention;
Fig. 3 is a conceptual diagram illustrating the composition of video display data formed by the compres-sion method of Fig. 2;
Fig. 4 is a conceptual diagram of a color space for explaining the color signal compression method of the invention;
Fig. 5 is a conceptual diagram for explaining an example of the color signal compression method of the invention;
Fig. 6 is a block diagram of an exemplary color video display circuit in the present invention;
Fig. 7 is a flow chart of a second method to encode the color signal compressed data according to the invention;
Fig. 8 is a conceptual diagram illustrating the composition of video display data formed by the compres-sion method of Figs. 7 and 10;
Fig. 9 is a waveform chart for explaining the effects of the invention;
Fig. 10 is a flow chart of a third method to encode the color signal compressed data according to the invention;
Fig. 11 is a flow chart of a fourth method to 1333~21 encode the color signal compressed data according to the invention;
Figs. 12A and 12B are conceptual diagrams illustrating the composition of video display data formed by the compression method of Fig. 11;
Figs. 13A and 13B are conceptual diagrams for explaining methods to decide a representative color as shown in Figs. 14 and 15;
Fig. 14 is a flow chart of a first embodiment to carry out the method of deciding a representative color; and Fig. 15 is a flow chart of a second embodiment to carry out the method of deciding a representative color.

DETAILED DESCRIPTION OF THE PREFERRED EMBODIMENTS:
Hereinafter the present invention will be described with reference to the accompanying drawings which show an exemplary color video display apparatus accomplished by applying a color video signal processing method and apparatus of the invention.
Fig. 1 is a block diagram of such color video display apparatus as a whole. Three primary color signals (red, green and blue) of a color image obtained from a 1333~21 color image input unit such as a video camera 1 are converted to digital primary color data R, G, B, each of which is composed of 8 bits, by analog-to-digital (A/D) converters 2R, 2G, 2B respectively, and then are stored temporarily in so-called video memories such as random access memories (RAMs) 3R, 3G, 3B each having a capacity of 2048 x 2048 pixels x 8 bits. Such memories 3R, 3G and 3B are connected to a main bus 4, to which a central processing unit (CPU) 5, a read only memory (ROM) 6, a RAM 7 and so forth are also connected. The color data compression is performed in the above consti-tution by the method to be described later, in such a manner that digital primary color data, where each color is composed of 8 bits (24 bits per pixel), are compressed to data dd of 8 bits per pixel. And such compressed data dd are recorded on a recording device 8 such as a hard disk or floppy disk, or are transmitted through a commu-nication apparatus such as a modulator-demodulator (MODEM) 9. A color video display circuit 10 is also connected to the main bus 4 and is supplied with the compressed color display data dd (8 bits per pixel). The color display data dd are restored in the color video display circuit 10 to the three primary color data R', G', B' of the former 24 bits per pixel where each color 133~21 is composed of 8 bits, and then are processed through digital-to-analog conversion to become analog primary color signals, which are fed to a color monitor CRT
(cathode-ray tube) 20.
Now a first exemplary embodiment to carry out the color data compression method will be described below with theference to the flow chart of Fig. 2. The digital primary color data R, G, B corresponding respec-tively to the saturations of the three primary colors (red, green, blue) of a color image are obtained by directly digitizing the original precompression data which are three primary color signals outputted from a video camera or the like, in a manner that the data of each color is composed of 8 bits per pixel (24 bits for the entire three primary colors). Such three primary color data R, G, B of 24 bits per pixel are compressed adaptively for the individual pixels of a color image as will be described later, so that display data dd of 8 bits per pixel is obtained. The 8-bit display data dd has a data format of Fig. 3, where three kinds of data dd(r), dd(g), dd(b) are set correspondingly to the individual primary colors, and selection of such data is determined adaptively in accordance with the absolute value of the difference between the digital primary color 1333~21 data of the adjacent pixels. With regard to the digital three primary color data R, G, B in Fig. 2, the primary color data of the _th pixel in one image are represented by Ri, Gi, Bi, respectively.
First in a step S101 shown in Fig. 2, initial values Ro, Go, Bo of the digital three primary color data R, G, B are set as Ro = 128, Go = 128, Bo = 128 And a count parameter i (corresponding to the pixel number) is set as 1 (i = 1). Subsequently in a step S102, there are calculated the absolute values ~R, ~G, aB of the respective differences between the primary color data Ri, Gi, Bi of the _th pixel and the primary color data Ri-l, Gi_l, Bi-l of the (i - l)th pixel which is immediately anterior to the _th pixel.
~R = ¦ Ri - Ri-l ¦
~G = ¦ Gi - Gi-l I
~B = ¦ Bi - Bi-l ¦
In a next step S103, the maximum of the above absolute values ~R, ~G, ~B is discriminated, and the digital primary color data corresponding to the color of such maximum value is compressed to the individual color saturation data of 6 to 7 bits in the following steps S104, S105 and S106, where an identification code of 1 1333~21 or 2 bits indicating the above color is added to produce 8-bit display data of the format shown in Fig. 3.
When the value ~R is judged to be maximum (~R > ~G, ~R ~ ~B) in the step S103, the process advances to the step S104 where the 8-bit digital red data Ri is compressed to 6-bit red saturation data r, and further two bits "10" of red identification code are added to the high-order side (MSB side) of the compressed data, thereby forming 8-bit display data dd(r) as a whole. In case the value ~G is judged to be maximum (~G > ~R, ~G ~ ~B) in the step S103, the process advances to the step S105 where the 8-bit digital green data Gi is compressed to 7-bit green saturation data g, and one bit "0" of green identi-fication code is added to the MSB side of the compressed data, thereby forming 8-bit display data dd(g). In case the value ~B is judged to be ~x;~um (~B ~ ~R, ~B > ~G) in the-step S103, the process advances to the step S106 where the 8-bit digital blue data Bi is compressed to 6-bit blue saturation data b, and two bits "11" of blue identification code are added to the MSB side of the compressed data, thereby forming 8-bit display data dd(b).
Now an embodiment for carrying out such color data compression method will be described in detail below.
Fundamentally, the data of 8 bits are converted to data of 1333~21 a smaller number of bits. The following is an exemplary case where mapping is performed by allocating m-bit compressed data correspondingly to the color saturation of the effective region in accordance with the color saturation distribution represented by the original data.
Therefore the display data dd(r) for instance has a sufficiently high quantization level although its color saturation information is composed merely of 6 bits.
~ore specifically, an example is illustrated in Fig. 4 where Rmax, Gmax, Bmax denote the respective maximum values of the primary color components of the colors used in, e.g. one frame of the color image being presently processed in the color space representable by the digital primary color data R, G, B each composed of 8 bits; and Rmin, Gmin, Bmin denote the respective minimum values of such primary color components. In this example, color data compression is performed by requantizing the primary color data R, G, B within such ranges to the primary color saturation data r, g, b composed of mr, mg, mb bits respectively in the above format. In calculation of the primary color saturation data r, g, b in such requantization, for instance, the saturation data r can be calculated as follows.
As shown in Fig. 5, since the original data R

1~33421 is composed of 8 bits, the gradation has a value in the color space ranging from 0 to 255, whereas the requantized data r has a value obtained by projecting the data R into a range of the gradation Rmin to Rmax. Therefore, when mr = 5~
r = (R - Rmin) x 25 / (Rmax - Rmin + 1) ...... (1) Similarly, when mg = 6 and mb = 5, g = (G - Gmin) x 26 / (Gmax - Gmin + 1) ...... (2) b = (B - Bmin) x 25 / (Bmax - Bmin + 1) ...... (3) Subsequently, in a next step S107, the color display data dd(r), dd(g) or dd(b) formed in the previous step S104, S105 or S106 is stored, as display data representing the _th pixel, in the _th address in a memory 7 serving as storage means. Then the process advances to a step S108 where the aforementioned count parameter i is incremented by 1. In a next step S109, a decision is made as to whether the above process has been executed completely with regard to the entire pixels. When the result of such decision is no, the process returns to the step S102; whereas if the result is yes, the entire process regarding one image is considered to be complete.
It follows, therefore, that the saturation information of the color having the maximum level change in comparison with the preceding pixel is stored as the data in the memory 7.
Fig. 6 shows an exemplary constitution of a color video display circuit 10 designed for displaying a color image by the use of such color display data compressed by the above-described method.
In Fig. 6, a display memory 11 has a capacity of (pixels of one image x 8) bits, and the color image display data dd (8 bits per pixel) is read out, synchro-nously with clock pulses CK, from the addresses corre-sponding to the horizontal and vertical scanning in a color display tube. Such 8-bit data dd is the one processed by the aforementioned data compression method.
Accordingly, relative bD each pixel data read out and transmitted from the memory 7 and the color having the maximum level change (color saturation change) in com-parison with the preceding pixel, the identification code and the saturation data of such color are stored as the display data dd in the dlsplay memory 11 of Fig.
6. The 8-bit display data dd read out from the memory 11 is fed as an address to the memory 12R, 12G or 12B
for the color look-up table.
Each of the memories 12R, 12G and 12B for the color look-up table has a capacity of 256 addresses accessible by the 8-bit display data dd, where each 133~21 address is composed of 9 bits. In the red memory 12R, predetermined conversion table data are previously stored in a manner that, when the two high-order or more significant bits A7 and A6 of the address A7 to A0 have become "10" of red identification code, the most significant bit D8 of the 9-bit output data D8 to D0 is turned to "1" signifying a renewal instruction, and the 8-bit digital red data R' is outputted to the address D7 to D0 in accordance with the saturation data r fed to the six more significant bits A5 to A0 of the address. More specifically, as the data content of the addresses 80H to BFH (where H denotes hexadecimal nota-tion) in the address space of the memory 12R, "1" is written in D8, and red data R' of the color space displayed actually in accordance with the aforementioned 6-bit saturation data r is written in D7 to D0. Mean-while in the green memory 12G, predetermined conversion table data are previously stored in the addresses 0 to 7FH
in a manner that, when the most significant bit A7 of the address A7 to A0 becomes "0" of green identification code, the most significant bit D8 of the 9-bit output data D8 to D0 is renewed to "1", and the 8-bit digital green data G' is outputted to the address D7 to D0 in accordance with the saturation data g which is fed to the 13~3 ;~ ~1 seven less significant bits A6 to AO of the address.
Further in the blue memory 12B, the conversion table data are previously stored in addresses COH to FFH in a manner that, when the two more significant bits A7 and A6 of the address A7 to AO become "11" of blue identifi-cation code, the most significant bit D8 of the 9-bit output data D8 to DO is renewed to "1", and the 8-bit digital blue data B ' is outputted to D7 to DO in accord-ance with the saturation data b fed to the six less significant bits A5 to AO of the address. Instead of such memories 12R, 12G, 12B for the color look-up table, other circuits may be employed for converting, by inverse arithmetic calculations based on Eqs. (1), (2) and (3), the compressed saturation data r, g, b into digital primary color data R ', G ', B ' each composed of 8 bits.
To latch circuits 13R, 13G, 13B, there are fed the eight less significant bit data D7 to DO of the memories 12R, 12G, 12B, i.e. the digital red data R', digital green data G ' and digital blue data B ' . Further-more, the most significant bit data D8 of the memories 12R, 12G, 12B are also fed as latch enable signals to the latch circuits 13R, 13G, 13B respectively. It follows, therefore, that the preceding data is held with 1333~21 regard to the color represented by "O" of the most significant bit data D8, and the data is renewed merely with regard to the color represented by "1".
In case the data dd obtained from the memory 11 is red data dd(r) for instance, since its two more significant bits are "10", only the most significant bit D8 of the output data from the memory 1 2R becomes "1"
and enables the latch circuit 13R to latch the eight less significant bits D7 to DO of the data, so that the red signal R is obtained from the D/A converter 14R.
Since the data dd are fed also to the memories 12G and 12B at this time, the data G' and B' are out-putted respectively from the memories 12G and 12B.
However, as the two more significant bits of the data dd are "10", the respective most significant bits D8 and D8 of the data outputted from the memories 12G and 12B do not become "1", and therefore the data G' and B' are not latched by the latch circuits 13G and 13B
respectively. Thus, the preceding data G' and B' latched anterior thereto are still held continuously.
Such state in the latch circuit 11 is retained until the next red data dd(r) is read out substantially from the memory 11, i.e. until the red saturation change becomes greater than the saturation change of some other 133~421 color.
In case the data dd read out from the memory 11 is green data dd(g) or blue data dd(b), the most significant bit D8 of the output data from the memory 12G or 12B becomes "1" in accordance with one more significant bit or two more significant bits thereof, so that the eight more significant bits D7 to DO of the data dd(g) or dd(b) are latched by the latch circuit 13G or 13B, whereby the green signal G or the blue signal B is taken out from the D/A converter 14G or 14B.
The operations of such latch circuits 13R, 13G
and 13B proceed in accordance with the aforementioned clock pulses CK, and the output data from the latch circuits 13R, 13G and 13B are converted respectively to red, green and blue analog primary color signals by the D/A converters 14R, 14G and 14B. Such analog primary color signals are then fed to a color monitor CRT 20 and so forth to display a color image.
According to the above-described technique where 6 bits are allocated to each of red and blue and 7 bits to green respectively, it becomes possible to execute substantially 19 bit color representation although the display data per pixel is composed merely of 8 bits, henae achieving simultaneous display of 1333~1 219 (= 524,288) colors in one image. Consequently, even a natural picture can be displayed with sufficient gradations to attain a satisfactory color image while maintaining desired naturalness and high picture quality.
Furthermore, with regard to any color having the maximum saturation change, the post-change display is executed with priority to eventually m; r~;m; ze deterioration of the resolution.
Besides the above, the color identification code is added merely to the data of the color having the m~X;mum saturation change, thereby eliminating the necessity of an increased capacity of the memory 11.
Moreover, any particular color compression circuit is not needed, and the required operation can be performed by mapping through the memories 12R to 12B.
In the above example, if the latch-enable signals from the more significant bits of the data dd to the latch circuits 13R to 13B are formed by the use of AND gates and inverters, the memories 12R to 12B may be such that each address thereof has a capacity of 8 bits. In this case, the sequence of the memories 12R to 12B and that of the latch circuits 13R to 13B can be reversed.
Although the above embodiment is so contrived 1333~1 that the data dd is once stored in the memory 7 in the step S107, the data may be directly stored in the memory 11. The procedure shown in Fig. 2 can be realized by either software or hardware.
Hereinafter a second embodiment for carrying out the method of color data compression will be described with reference to a flow chart of Fig. 7. However, the explanation is simplified with respect to the steps corresponding to those in the aforementioned first embodi-ment of Fig. 2.
The 8-bit display data dd processed by such compression method has a data format shown in Fig. 8.
First in a step S201, initial values are set as Ro = 128, Co = 128, Bo = 128 Yo = 0.3 Ro + 0.59 Go + 0.11 Bo, i = 1 In a next step S202, lllmin~nce data Yi of the _th pixel is calculated from the data Ri to Bi as follows:
Yi = 0.3 Ri + 0.59 Gi + 0.11 Bi There are also calculated the absolute values ~R, ~G and ~Y of the respective differences between the data Ri to Yi of the _th pixel and the data Ri_l to Yi_l of the preceding pixel.
~R = ¦ Ri - Ri-l ¦
~G = ¦ Gi - Gi-l I

133~4~1 ~B = ¦ Bi - Bi-l ~y = I Yi - Yi-l I
Subsequently in a step S203, the maximum of the absolute values ~R to ~Y is discriminated.
When the value ~R is judged to be maximum (~R > QG, ~R > ~B), the process advances to a step S204 where the 8-bit digital red data Ri is compressed to 6-bit red saturation data r and, as shown in Fig. 8, two bits "00" of red identification code are added to the MSB side of the data r, thereby forming 8-bit display data dd(r) as a whole.
In case the value ~G is judged to be maximum (~G > ~R, ~G > ~B, ~G > ~Y) in a step S203, the process advances to a step S205 where the 8-bit digital green data Gi is compressed to 6-bit green saturation data g, and further two bits "01" of green identification code are added to the MSB side of the data, thereby forming 8-bit display data dd(g) as a whole.
If the value ~B is judged to be maximum in the step S203, the process advances to a step S206 where the 8-bit digital blue data Bi is compressed to 6-bit blue saturation data b, and two bits "10" of blue identifica-tion code are added to the MSB side of the compressed data, thereby forming 8-bit display data dd(b) as a whole.

_ 24 1333~1 Furthermore, in another case where the value aY is judged to be maximum in the step S203, the process advances to a step S207 where the 8-bit digital luminance data Yi is compressed to 6-bit luminance data y, and two bits "11" of luminance identification code are added to the MSB side of the compressed data, thereby forming 8-bit display data dd(y) as a whole. The above compres-sion is performed in the same manner as in.the aforemen-tioned first embodiment.
Subsequently in a next step S208, the data dd(r), dd(g), dd(b) or dd(y) formed in one of the previous steps S204 to S207 is stored as _th pixel display data in the _th address of the memory 7. Thereafter in a step S209, the count parameter i is incremented by 1 so as to process the next pixel. In a step S210, a decision is made as to whether the above process has been completed with regard to the entire pixels and, if any pixel is left nonprocessed, the procedure posterior to the step S202 is executed repeatedly.
Thus, there is stored in the memory 7 the satura-tion information as the data dd relative to the color or luminance having the maximum level change in comparison with the preceding pixel.
In the first and second embodiments for carrying 1333~21 out the display data forming method mentioned above, the operations shown in Figs. 2 and 7 may be performed con-tinuously for each image. However, considering the horizontal correlation, it is desired that initialization be executed per line.
In the steps S102 and S202, the present pixel data such as Ri is compared with the preceding pixel data Ri_l. However, for further reducing the error, the data Ri may be compared with the original pixel data Rx corre-sponding to the latest display data dd(r) stored in the memory 7 in the step S107 or S207. And similar comparison may be performed with respect to any of the other original data Gx, Bx and Yx corresponding respectively to the latest display data dd(g), dd(b) and dd(y).
It is to be understood that the above-described method for forming the compressed data is merely an example, and some other compressed data may be formed on condition that m-bit data is obtainable by mapping the original data.
The color display circuit 10 for displaying a color image with the color display data compressed by the method of the second embodiment is fundamentally the same as the aforementioned circuit shown in Fig. 6. However, in the memory 12R, predetermined data are previously 1333~2l stored in a manner that when the two high-order or more significant bits A7 and A6 of the address A7 to AO have become "OO" of red identification code or "11" of luminance identification code, the most significant bit D8 of the data D8 to DO is turned to "1", and also that the data of the red component level of the color to be displayed is outputted as low-order or less significant bits D7 to DO
(red data R'). That is, in the memory 12R are stored "1"
(D8) and the red data Rl (D7 to DO) in the address OOH to 3FH and the address COH to FFH (where H denotes hexadecimal notation).
Meanwhile in the memory 12G, predetermined data are previously stored in a manner that when the two more significant bits A7 and A6 of the address A7 to AO have become "Ol" of green identification code or "11" of luminance identification code, the most significant bit D8 of the data D8 to DO is turned to "1", and also that the data of the green component level of the color to be dis-played is outputted as less significant bits D7 to DO
(green data G'). That is, in the memory 12G are stored "1" (D8) and the green data G' (D7 to DO) in the address 4OH to 7FH and the address COH to FFH. Further in the memory 12B, predetermined data are previously stored in a manner that when the two more significant bits A7 and A6 1333~21 of the address A7 to A0 have become "10" of blue identi-fication code or "11" of luminance identification code, the most significant bit D8 of the data D8 to D0 is turned to "1", and also that the data of the blue compo-nent level of the color to be displayed is outputted as less significant bits D7 to D0 (blue data B'). That is, in the memory 12B are stored "1" (D8) and the blue data B' (D7 to D0) in the address 80H to BFH and the address COH to FFH.
When the two more significant bits of the data dd are "11", the addresses COH to FFH of the memories 12R
to 12B are designated simultaneously, so that the data R' to B' are outputted simultaneously from the memories 12R
to 12B. The values of such output data obtained in this stage are set at the following ratio:
R' : G' : B' = 0.3 : 0.59 : 0.11 ....... (1) However, the absolu ~ values of the data R', G' and B' conform to the six less significant bits of the data dd.
The operations of the latch circuits 13R to 13B
are fundamentally the same as those in the aforementioned first embodiment. When the data dd from the memory 11 is the lllm;n~nce data dd(y), the data R' to B' are latched simultaneously in the latch circuits 13R to 13B. In this stage, since the data R' to B' are set at the ratio of 1333~21 Eq. (1), the levels of the signals R to B obtained from the D/A converters 14R to 14B alsc come to have the same ratio as R : G : B = 0.3 : 0.59 : 0.11 And the signals R to B serve as luminance signals (bright-ness information).
Relative to any color having great saturation change, its post-change display is executed with priority.
In addition, an idea of the luminance is adopted so that, when there exists great lllm;nance change in the original data, the signals R to B are simultaneously changed correspondingly to 18 bits in accordance with the iden-tification code "11" to change the lllm; n~nce of the display image, thereby achieving a remarkably high resolution.
Due to adoption of such luminance idea, there occurs no phenomenon that, for instance, the boundary between black and white portions is colored. That is, in the boundary between black and white portions, the red to blue data (Ri to Bi) change simultaneously to great extent as shown in Fig. 9. And if none of the signals Yi and ~Y
is employed, the maximum of the values aR, aG and ~B is discriminated in the step S203. For example, under the conditions of ~G > ~R and ~G > ~B, the value ~G is judged ~9 13~3~21 to be maximum and, in the step S205, the data dd(g) is regarded as the data dd of the -th pixel. Consequently, at the display time, the red and blue data of the _th pixel are considered to represent the saturation of the (i - l)th pixel which is immediately anterior to the _th pixel, while the green data is considered to represent the post-change saturation. Therefore in the _th pixel, the green saturation becomes greater than the red and blue saturations, so that the pixel is colored in green.
And the same phenomenon occurs with respect to the next (i + l)th pixel as well.
However, according to the present invention where the signals Yi and ~Y are formed with adoption of the luminance idea, the saturations of the individual colors in the _th pixel are simultaneously changed to consequently avert undesired coloring.
Now a third embodiment for carrying out the method of color data compression will be described below with reference to a flow chart of Fig. 10. The 8-bit display data dd compressed by such method has a data format shown in Fig. 8.
First in a step S301, initial values are set as Rx = 0.3 Ri + 0.59 Gi + 0.11 Bi Gx = Rx, Bx = Rx, i = 1 1333~2l In the above, the values RX, Gx and Bx are replaceable with the values Ri, Gi and Bi respectively, as described complementally in connection with the second embodiment, when the data Ri, Gi and Bi satisfy the undermentioned predetermined conditions.
In a next step S302, the luminance data Yi of the _th pixel is calculated from the data Ri to Bi as Yi = 0.3 Ri + O . 59 Gi ~ 0.11 Bi And there are also calculated the absolute values ~R to ~B of the differences between the data Ri to Bi and the data RX to Bx, and the absolute values ~Yr to ~Yb of the differences between the data Yi and the data Ri to Bi:
~R = ¦ Ri - Rx ¦
~G = ¦ Gi - Gx ¦
QB = ¦ Bi Bx ¦
~Yr = ¦ Yi - Ri ¦

g = ¦ Yi - Gi ¦
~Yb = ¦ Yi - Bi ¦
Subsequently in a step S303, greater one of the differences ~G and ~B is selected as a difference ~S;
greater one of the differences ~R and ~B is selected as a difference ~T; greater one of the differences ~R and ~G
is selected as a difference ~U; and the greatest one of the differences ~Yr to ~Yb is selected as a difference ~V.

1333~2l In a next step S304, the differences ~S to ~V are compared with one another so that the minimum is discriminated among such differences.
If the value ~S is judged to be minimum (~S > ~T, as > ~u, ~s > QV) out of such differences QS to ~V, it signifies that, when the red data R is renewed to Ri, the greater one of the residual errors ~G and ~B caused in the green and blue data is smaller than the residual error caused with renewal of any other color data than the red data. Therefore the data Gi and Bi in this stage can be replaced with the data Gx and Bx of the preceding pixel, hence eliminating the necessity of preparatorily storing the green and blue data Gi and Bi of the _th pixel in the memory. Such technique enables preparatory storage of the red data Ri of the _th pixel in the memory, and the requirement is renewal of the data Rx alone to the data Ri.
The same process as the above may be executed when any of the differences QT to ~V is m;n;mum.
If the difference ~S is judged to be minimum, the process advances to a step S305, where Rx is replaced with Ri, and the 8-bit original data Ri is converted to 6-bit compressed data r. Furthermore, two bits "00" of red identification code are added to the high-order side (MSB
side) of the compressed data r as shown in Fig. 8, thereby 1333~21 forming 8-bit display data dd(r) as a whole.
When the difference ~T is judged to be minimum in the step S304, the process advances to a step S306, where Gx is replaced with Gi, and the 8Obit original data Gi is converted to 6-bit compressed data g. Then two bits "01" of green identification code "01" are added to the MSB side of the compressed data as shown in Fig. 8, thereby forming 8-bit display data dd(g).
If the difference ~U is judged to be minimum in the step S304, the process advances to a step S307, where Bx is replaced with Bi, and the 8-bit original data Bi is converted to 6-bit compressed data b. Furthermore, two bits "10" of blue identification code are added to the MSB side of such compressed data as shown in Fig. 8, thereby forming 8-bit display data dd(b) as a whole.
In case the difference ~V is judged to be minimum in the step S303, the process advances to a step S308, where Rx, Gx and Bx are replaced with Ri, Gi and Bi respectively, and the 8-bit original data Yi is converted to 6-bit compressed data y. Then two bits "11" of luminance identification code are added to the MSB side of the compressed data as shown in Fig. 8, thereby form-ing 8-bit display data dd(y) as a whole.
Subsequently in a next step S309, the data dd(r), dd(g), dd(b) or dd(y) formed in one of the previous steps 1333~21 S305 to S308 is stored as _th pixel display data in the _th address of the memory serving as storage means.
Thereafter in a step S310, the count parameter i is incremented by 1 so as to process the next pixel. In a step S311, a decision is made as to whether the above process has been completely executed with regard to the entire pixels and, if any pixel is left nonprocessed, the procedure posterior to the step S302 is executed repeatedly.
Thus, there is stored in the memory the gradation information as data dd relative to the color or luminance having the maximum level change in comparison with the preceding pixel.
The color display circuit 10 for displaying a color image with the color display data compressed by the method of the third embodiment may be fundamentally the same as the aforementioned circuit shown in Fig. 6, and its operation is the same as that of the second embodi-ment.
However, in the addresses COH to FFH of the memories 12R to 12B, the same data is stored in each of the memories 12R, 12G and 12B.
When the data dd from the memory 11 is the lllm;n~nce data dd(y), since its two more significant bits are "11", mutually equal data R', G' and B' are outputted from the entire memories 12R to 12B, and the respective most significant bits D8 to D0 are turned to "1". There-fore, the data R', G' and B' are latched simultaneously in the latch circuits 13R, 13G and 13B, and signals R to B having luminance information (brightness information) are obtained from the A/D converters 14R to 14B.
Consequently, relative to any color having great gradation change, its post-change display is executed with priority. In addition, an idea of the luminance is adopted so that, when there exists great lllm;n~nce change in the original data, the signals R to B are simultane-ously changed correspondingly to 18 bits in accordance with the identification code "11" to change the luminance of the display image, thereby achieving a remarkably high resolution.
In case none of the signals Yr to ~Yb and ~V is employed, the minimum of the values ~R, ~G and ~B is discriminated in the step S304. For example, under the conditions of ~S < aT and ~S < ~U, the value ~S is judged to be ~;n;mum. Then, in the step S305, the data dd(r) is regarded as the data of the _th pixel. Consequently, at the display time, the green and blue data of the _th pixel are considered to have the level of the preceding pixel such as the (i - l)th pixel, while the red data is considered to have the post-change level. Accordingly, in the -th pixel, the red level becomes higher than the green and blue levels, so that the pixel is colored in red.
However, according to the present invention where the signals ~Yr to QYb and QV are formed with adoption of the luminance idea, the levels of the individ-ual signals R to B in the _th pixel are simultaneously changed to consequently avert an undesired phenomenon that the boundary between black and white portions is colored. It is to be understood that, for simplifying the circuit configuration, the above-described process relative to the luminance may be omitted as well.
Now a fourth embodiment for carrying out the method of color data compression will be described below with reference to Fig. 11. The data format in this example is shown in Fig. 12A, where 8 bits of display data dd are allocated to one pixel as in the aforemen-tioned first through third embodiment. Fundamentally a control code is composed of 3 high-order bits, and color saturation data is composed of 5 low-order bits. When the 3 high-order bits of each of the display data dd(r), dd(g) and dd(b) are "000", the 5 low-order bits of the 8-bit red 133~21 data Ri compressed to saturation data r. Similarly, when the 3 high-order bits are "001" and "010", the 5 low-order bits of the 8-bit green data Gi and those of the 8-bit blue data Bi are compressed to saturation data g and b, respectively.
In case the 3 high-order bits are "011" to "111", the 5 low-order bits represent specific display data dd(j) (where j = 1 to 5) as will be described later. As illustrated in Fig. 4, a region 30 having a distribution of signals R to B is sliced to form 32 layers G'l to G'32 of, e.g. 5 bits (mg = 5) in parallel with the R-B plane or in the direction of G, and further each Gk of such layers G'l to G'32 is divided into 32 x 32 subregions 31 of, e.g. 5 bits (mr = 5, mb = 5) relative to red signal R and blue signal B. It follows that the red, green and blue gradation data r, g and b belong to any of the subregions 31.
The appearance frequency distribution of the sub-regions 31 is obtained by using the entire pixel data of the image. In this case, any subregion of a high frequency denotes the color appearing frequently in the image.
Five highest appearance frequency subregions 31 are selected out of each layer G'k of the green layers G'l to G'32, and the colors of such five subregions 31 are 1333~21 chosen as representative colors Pj (where j = 1 to 5).
The data of the representative colors Pj in each layer Gk are registered as display data dd(j) together with 5-bit color saturation data obtained by compressing the green data of that layer correspondingly to the control code "011" to "111". Since there are a total of 32 green layers G'l to G'32 each having 5 representative colors Pj, it follows that 160 representative colors are existent in all. The red, green and blue gradations in the representative colors Pj are regarded as, e.g. the center values of the subregions.
In coloring one pixel, a calculation is executed to decide which data of the red, green, blue and repre-sentative colors needs to be renewed (while the data of the remaining colors are not renewed) between such pixel and the preceding pixel to achieve the most adequate coloring for the original image, and the color data selected to attain the most adequate coloring is stored as display data dd for the pixel.
As shown in Fig. 11, the pixel data at the top of each horizontal line is initialized in steps S401 through S405. To perform this operation, in the first step S401, the pixel data Ro to Bo at the top of a hori-zontal line are read out from memories 3R to 3B. In the 133~21 next step S402, the data Go out of the entire data Ro to Bo read out in the step S401 is requantized with 5 bits on the basis of the dynamic range of the region shown in Fig. 4, thereby forming data g.
Subsequently in the step S403, there are read out the data Rj *, G*, Bj * (where j = 1 to 5) of the representative colors Pj in the layer including the data g out of the layers G'l to G'32 in the region 30. And the color intrablock distance Lj between such data Rj*, G*, Bj* and the data Ro, Go, Bo read out in the step S401 is calculated in the step S404 according to the following equation:
Lj = ~(Ro - Rj*) 2 + (Go - Gj*l2 + (Bo - Bj*) 2 where j = 1 to 5. Subsequently the minimum value imin of the distance Lj is obtained in the step S405, and the control code and the gradation data are set as follows in the step S406.
control code b7 to b5 = imin + 2 Gradation data b4 to bo = g In the control code, addition of 2 as Jmin + 2 is based on the fact that 0 to 2 of the control code are allocated to red to blue, and 3 to 7 thereof are allocated to the representative colors Pj. In a step S407, the replace-ment values Rx, Gx, Bx are obtained as - 3q 1333~21 Rx = Rj*, Gx = G*, Bx = Bj*
The data of each pixel are produced in steps S408 to S419 in the following manner. First in the step S408, the next pixel data Ri to Bi are read out from the memories 3R to 3B, and the differences ~R, ~G, ~B between such next pixel data Ti to Bi and the replacement values Rx to Bx (in this example, the values obtained in step S407) are calculated as ~R = ¦ Ri - Rx ¦
~G = ¦ Gi - Gx ¦
~B = ¦ Bi - Bx ¦
And in a next step S409, the maximum value of such differences ~R to ~B is discriminated.
When the difference ~R is ]udged to be maximum, the process advances to a step S410, where the data Ri out of the entire data Ri to Bi read out in the step S408 is quantized with 5 bits on the basis of the dynamic range of the region 30 as in the step S402, thereby forming data r as (Ri - Rmin) x 25 Rmax - Rmin + 1 nd the control code and the gradation data are set as Control code = "000"
Gradation data = r Furthermore, the replacement value Rx is calculated as Rx = (r + 1/2) (Rmax - Rmin + 1) / 25 + Rmin If the difference ~G or AB is judged to be m~X;mum in the step S409, the foregoing process of the step S410 is executed similarly in a step S411 or S412:

= (Gi - Gmin) x 25 Gmax - Gmin + 1 Control code = "001"
Gradation data = g Gx = (g + 1/2) (Gmax - Gmin + 1) / 25 + Gmin or b = (Bi - Bmin) x 25 BmaX ~ Bmin + 1 Control code = "010"
Gradation data = b Bx = (b + 1/2) (Bmax - Bmin + 1) / 25 + Bmin Subsequently to the step S410, S411 or S412, the square of the residual error ~o: is calculated in a step S413 as ~ o2 = (Ri - Rx)2 + (Gi - Gx)2 + (Bi - Bx)2 And in a step S414, the data Gi out of the present pixel data Ri to Bi is quantized with 5 bits on the basis of the dynamic range of the region 30 similarly to the afore-mentioned process in the step S402, thereby forming the data g.

g = (Gi - Gmin) x 25 GmaX - Gmin + 1 In a next step S415, there is calculated the square of the residual error aj caused when using 5 representative colors Pj (where j = 1 to 5) in the layer including the data g out of the layers G'l to G'32 in the region 30.
aj2 = (Ri - Rj*)2 + (Gi - G*)2 + (Bi - Bj*)2 where j = 1 to 5. Then in a step S416, the values aO2 obtained in the step S413 are compared with the values aj2 obtained in the step S415, and a check is executed as to whether any of the values ao2 is minimum.
In case none of the values ao2 is minimum, that is, when one of the values aj2 is minimum, the process advances to a step S417 where the control code and the gradation data are set as follows with respect to the jminth representative color Pj having the m; n;mum value aj .
Control code = jmin + 2 Gradation data = g Subsequently in a step S418, the replacement values Rx to Bx are obtained for the next pixel as Rx = Rj*, Gx = G*, Bx = Bj*
And then the process advances to a step S419.

When the value ~o2 is judged to be minimum in the step S416, the process advances to a step S419 while skipping over steps S417 and S418.
In the step S419, the data dd obtained in one of the steps S410 to S412 or in the step S417 is written as the present pixel data in the memory.
Thereafter the count parameter i is incremented by 1 in a step S420, and then a check is executed in a step S421 as to whether the above-described data processing procedure has been completed or not with regard to the entire pixels of the present horizontal line. And if the result of such check is no, the process returns to the step S408, and the same procedure is repeated as mentioned for the next pixel.
Upon completion of the data processing procedure for the entire pixels of the present horizontal line, the process advances from the step S421 to a next step S422, where a check is executed as to whether the aforementioned data processing procedure has been completed or not with regard to the entire horizontal lines. And if the result is no, the process returns to the step S401 and thereafter the foregoing procedure is repeated as mentioned for the pixel of the next horizontal line.
This program is terminated upon completion of ~ ~ _ the data processing procedure for the pixels of the entire horizontal lines, whereby the data dd compressed to 8 bits per pixel is stored in a floppy disk 8 or the like. The video display circuit 10 for displaying the data compressed in the above fourth embodiment is con-stituted as follows. In Fig. 6, predetermined data are previously stored in the memory 12R in a manner that, when the three high-order or more significant bits A7 to A5 of the address A7 to AO have become "000" of the red control code or "011" to "111" of the representative color control code, the most significant bit D8 of the data D6 to DO iS turned to "1", and also that the data of the red component level of the color to be displayed is outputted as low-order or less significant bits D7 to DO (red data R').
Meanwhile in the memory 12G, predetermined data are previously stored in a manner that when the three more significant bits A7 to A5 of the address A7 to AO
have become "001" of the green control code or "011" to "111" of the representative color control code, the most significant bit D8 of the data D8 to DO is turned to "1", and also that the data of the green component level of the color to be displayed is outputted as less significant bits D7 to DO (green data G').

1333~21 Further in the memory 12B, predetermined data are previously stored in a manner that when the three more significant bits A7 to A0 of the address A7 to A0 have become "010" of the blue control code or "011" to "111" of the representative color control code, the most significant bit D8 of the data D8 to D0 is turned to "1", and also that the data of the blue component level of the color to be displayed is outputted as less significant bits D7 to D0 (blue data B').
Accordingly, discrimination is executed per pixel as to whether the data dd is the red, green, blue or representative color data by the control code, so that the color data of each pixel having the maximum gradation change in comparison with the preceding pixel is discriminated. Then the data in one of the latch circuits 13R to 13B is renewed in accordance with the color having the maximum gradation change, and the signal R to B are taken out.
Thus the three primary color signals R to B
can be obtained. In the present invention where 5 bits are allocated to each of red, green and blue data, there are also provided a total of 160 representative colors.
It follows that one color is expressed with 15 bits to consequently realize simultaneous production of 215 =

1333~21 32,768 colors. Therefore, even a natural image can be displayed with sufficient gradations while its naturalness and lustrous colors are maintained.
Furthermore, due to the adaptive process with representative colors, there occurs no color blur to eventually avert deterioration of the resolution. Besides the above, since the data in the memories 12R to 12B con-stituting the color look-up table are formed in view of statistic characteristics, a high efficiency is attainable in the color compression. And desired color compression can be achieved merely by the technique of mapping without the necessity of any particular additional circuit.
Fig. 12B shows another exemplary composition of video display data dd. In this example, 6 bits are allocated to each of red, green and blue data, so that a total of 218 = 262,144 colors can be produced and used in combination with a total of 24 layers x 4 = 64 representa-tive colors.
In deciding the representative colors by the method mentioned, there arises a problem of color blur in image portions such as edges if the appearance frequencies of the colors are not uniform in the image.
Suppose now that the color appearance frequency distribution of red to blue in one layer G'k out of the 13334~1 green layers T'l to G'32 is such as illustrated in Fig.
13A, where the curves are drawn, like contour lines in a map, by connecting the points (colors) of the same appearance frequency in the R-B coordinate system. It signifies that a point PA (i.e. color at point PA) appears most frequently.
In view of such state where the appearance frequency is concentrated at the point PA, if five points of higher appearance frequencies are selected as repre-sentative points, and the colors at such selected points are determined to be representative colors Pj (where j =
1 to 5) in the layer G'k, then the five representative points providing such representative colors Pj are con-centrated in the vicinity of the point PA, so that preparation of the five representative points or colors Pj is rendered ineffective to eventually bring about deterioration of the efficiency.
Therefore, in the present invention, the follow-ing process is executed with regard to each green layer G'k. Relative to the highest appearance frequency Nmax at the point PA where the color appearance is most frequent, a value Nx is calculated as Nx = Nmax / M
where M is an integer greater than 1, such as 100. Then _ 47 any appearance frequencies under the value Nx are cut off, so that a distribution of the appearance frequencies higher than the value Nx is obtained as illustrated in Fig. 13B. In this case, the appearance frequencies thus cut off are less than l/M of the highest appearance frequency Nmax, e.g., less than 1~ when M is 100, whereby the influence of such cut-off operation is rendered negligible.
In such cut-off appearance frequency distribution, the point PA is set as 1st representative point, and then a point PB farthest from the point PA is selected as a 2nd representative point. Subsequently a point Pc farthest from both points PA and PB is selected as a 3rd representa-tive point, and thereafter points PD and PE farthest from all of the representative points already set are selected as 4th and 5th representative points.
Then, data of the 1st to 5th representative colors Pl to P5 are obtained from the 1st to 5th repre-sentative points thus selected.
The 1st to 5th representative points are set at proper positions of which appearance frequencies are higher than a certain degree and which are sufficiently spread in the space.
The program shown in Fig. 14 is executed by a CPU

5 to decide representative colors Pj in the green layers G'l to G'32.
When the program is started, first in a step S501, the m; n;mum values Rmin to Bmin and the maximum values Rmas to Bmax are detected from the one-frame data of the memories 3R to 3B. And in a next step S502, the data R to B are requantized with 5 bits on the basis of the dynamic range of the region 30 as = (R - Rmin) x 25 Rmax - Rmin + 1 g = (G - Gmin) x 25 Gmax - Gmin + 1 (B - Bmin) x 25 Bmax - Bmin + 1 where the count parameter k is 1. Subsequently in a step S503, there is calculated per color the frequency distribution of the data r to b, i.e., the appearance frequency nk f each color (in r-b coordinate system) in the green layers G'k (where k = 1 to 32).
In a next step S504, there are found the highest appearance frequency Nk in each of the green layers G'k, and also the coordinates (rl, bl corresponding to point PA) where such value Nk is obtained. And in a step S505, any appearance frequencies under the value nk are cut off.

13334~1 When M = 100 for example, the appearance frequency nk defined as nk ~ Nk/100 is regarded as nk = -Subsequently in a step S506, there is calculatedthe distance D2 from the highest appearance frequency point (rl, bl) in comparison with each appearance fre-quency point (rn, bn) where nk ~ -D2 = ~(rn - rl)2 + (bn - bl)2 In a step S507 are found the longest distance D2 and the coordinates (r2, b2) corresponding to a point PB where the distance D2 is m~X; mum .
In a step S508, the value j is set to 3 for a point Pc. Then in a step S509, there is calculated each distance Dj from the points (rl, bI) - (rj-l, bj-l) in comparison with the points (rn, bn) where the appearance frequency nk = 0. And in a next step S510, there are found the longest distance Dj and the coordinates (rj, bj) corresponding to a ith point where the distance Dj is maximum.
Subsequently the value j is incremented by 1 in a step S511, and then a check is executed as to whether j _ 5 or not in a step S513. In case j- _ 5, the process returns to the step S509. Accordingly, in the steps S509 to S511 are calculated the 4th and 5th points (r4, b4), (r5, b5) and the distances D4, D5 therefrom.

13334~1 Since j ~ 6 in a next step S512, the process advances to a step S513, where color levels Rj*, G*, Bj*
are allocated to the representative points (rj, bj) (where j = 1 to 5) in the green layers G'k:
Rj* = (rj + 1/2) x (Rmax - Rmin + 1) / 25 + Rmin Gx = (g + 1/2) x (Gmax - Gmin + 1) / 25 + Gmin Bj* = (bj + 1/2) x (Bmax - Bmin + 1) / 25 + Bmin The count parameter k is incremented in a step S514. And then the foregoing operation is repeated with the parameter k changed in sequence until the program is executed completely.
According to the present invention, the repre-sentative points (rj, bj) in each green layer G'k are thus determined. In particular, such representative points are sufficiently spaced apart from one another within each green layer G'k and are so selected that the color appearance frequencies thereat are higher than a certain degree in each layer G'k, whereby selection of the representative points and colors is rendered achiev-able with a high efficiency. And even in any image where the color appearance frequencies are locally con-centrated, it is still possible to accomplish adequate selection of representative colors spread spatially.
Furthermore, since the process required for such 1333~21 selection is realizable merely by repetition of routines, fast operations can be performed, and execution by hard-ware is also effected simply and quickly.
In the above-described embodiments, the repre-sentative points and colors are determined in accordance with the appearance frequency distribution within each of the green layers G'k. However, it is to be understood that three-dimensional determination of representative points and colors can also be attained with reference to the appearance frequency distributions in the other two green layers above and below the desired layer to be processed.
In another embodiment of the present invention, adaptive mapping of representative colors is executed by first finding a pixel of an edge portion or the like where the difference between such pixel and the adjacent pixel becomes greater than a predetermined threshold value, then calculating the appearance frequencies of the colors corresponding to the mutually adjacent pixels and appearing in each subdivision of the color space, and using the selected colors of higher appearance fre-quencies with priority.
Now the method of selecting such representative colors will be described below with reference to Fig. 15.

1333~21 First in a step S601, relative to the colors used in one frame of a color image to be processed and developed in the R, C, B color spaces as shown in Fig.
4, the maximum and minimum values (Rmax, Rmin, Gmax, Gmin, Bmax, Bmin) along the respective axes of the primary color data are calculated.
In a next step S602, a decision is made as to whether the present pixel is at the top of a horizontal line. If the result of such decision is yes, the process advances to a step S607; and when the result is no, the process may advance to a step S605 in Fig. 5. However, for reducing the amount of calculation in this embodiment, the process advances to steps S603 and S604 so as to narrow down the edge pixel.
In the step S603, there are obtained the absolute values of the respective differences between the digital primary color data Ri, Gi, Bi of the present pixel (gen-erally _th pixel) and the primary color data Ri-l, Gi-l, Bi-l of the preceding (1 - i)th pixel adjacent to the _th pixel, and then the sum ~i' of such absolute values is calculated as ~ Ri - Ri-l ¦ + ¦ Gi - Gi-l ¦ + ¦ Bi - Bi-l ¦
And in the next step S604, a decision is made as to whether the sum ~i' is greater than a specific threshold value Eth' (i.e. ~i' _ Eth') predetermined to discriminate the edge. The process advances to a step S605 if the result of such decision is yes, or to a step S609 in the case of no. In the steps S603 and S604, there is detected, in one frame of the color image, a pixel of edge or con-tour portion having a great signal change in comparison with the adjacent pixel. The pixel detected in this step represents a signal change along the scanning direction in display of the image, e.g., an edge pixel in the horizontal direction.
Subsequently in a step S605, there is calculated the intermediate value (or 2nd value) ~i out of the absolute values of the respective differences between the digital primary color data Ri, Gi, Bi of the present _th pixel and the primary color data Ti-l, Gi-l, Bi-l of the preceding (i - l)th pixel adjacent to the _th pixel:
~i = Intermediate value {¦ Ri - Ri-l ¦, ¦ Gi - Gi-l ¦, ¦ Bi - Bi-l ¦}
And in a next step S606, a decision is made as to whether the intermediate value ~i is greater than a specific threshold value Eth (i.e. ~i 2 Eth) predetermined to dis-criminate the edge. Then the process advances to a step S607 if the result of such decision is yes, or to a step S609 in the case of no.

13~3421 In the steps S605 and S606, relative to the compressed color display data with color identification code added to the compressed saturation data of the color having the maximum saturation change between mutually adjacent pixels as in the adaptive mapping described previously, there is detected the maximum difference between the color data obtained by restoring the compressed color display data and the color data of the original color image. Since the maximum of the absolute values of the respective differences between the above-described primary color data is employed as the saturation data out of the compressed data in the adaptlve mapping, substantially no contribution is made to the difference between the restored data and the original color data. Considering such fact, the 2nd value ~i is compared with the predetermined threshold value Eth to detect the pixels where the difference between the restored data and the original color data is great. The pixels thus detected represent the dete-riorated portion in the adaptive mapping more exactly than the pixels detected in the steps S603 and S604, but discrimination between the.great and small differential values is needed to consequently necessitate arithmetic operation similar to the adaptive mapping. Meanwhile in the steps S603 and S604, the process is merely to find the sum of the absolute values of the respective differ-ences between the mutually adjacent pixels with regard to the digital primary color data R, G, B of the original color image, and the discrimination between the great and small differential values is not needed to eventually simplify the calculation. Accordingly, the overall processing speed can be increased by previously narrow-ing down, in the steps S603 and S604, the pixels for arithmetic operation in the steps S605 and S606, and thereby reducing the amount of required calculation in the steps S605 and S606. It is a matter of course that the decision levels in the steps S606 and S604 can be controLled by properly adjusting the specific threshold values Eth and Eth'. Either of the steps S603 and S604, or either of the steps S605 and S606, may be omitted as well.
In the step S607, each range between the maximum and m; n;mum values of the color space per primary color axis is requantized with 5 bits, and the digital primary color data Ri, Gi, Bi of the 1th pixel are compressed to color saturation data ri, gi, bi each composed of 5 bits.
Specifically, the calculations for such requantization are executed as ri = (Ri - Rmin) x 25 / (Rmax - Rmin + 1) gi = (Gi - Gmin) x 25 / (Gmax - Gmin + 1) Bi = (Bi - Bmin) x 25 / (Bmax - Bmin + 1) Subsequently in a step S608 are counted the color appearance frequencies f (r, g, b) in the color space compressed along each axis of the requantized color saturation data r, g, b. In this step, the range between the maximum value Gmax and the minimum value Gmin on the G-axis in the color space shown in Fig. 4 is requantized to obtain 32 divisions each composed of 5 bits, thereby forming 32 equivalent g-planes in the individual quanti-zation step. Each of the ranges Rmax-Rmin and Bmax-Bmin is requantized with 5 bits to such 32 g-planes, then each of the g-planes is divided into 32 x 32 - 1024 subregions, and the color appearance frequencies f (r, g, b) in the entire pixels of one frame are counted per subregion.
In a step S609 where the next (i + l)th pixel adjacent to the present pixel is processed, the count parameter i is incremented by 1 as (i = i + 1).
Subsequently in a step S610, a decision is made as to whether the above procedure has been completed or not with regard to the entire pixels in one frame of the color image. And the process advances to a step S601 if the result of such decision is yes, or returns to the 13~3421 step S602 in the case of no.
In a step S611, there are obtained the maximum value of the appearance frequencies f (r, g, b) per g-plane shown in Fig. 4, and also the values of r and b (coordinates on the r-axis and b-axis in the compressed color space). On the g-plane corresponding to one address k (k = 0 to 31) on the g-axis in the compressed color space, the color appearance frequency f (r, g, b) can be expressed as a function f (r, k, b) or fg=k (r b where the value of g is fixed to the above k. If the function with such parameters r and b is plotted by means of contour lines (appearance frequency lines) as illus-trated in Fig. 13A, the calculation of the maximum appearance frequency corresponds to calculation of the coordinates (rA, bA)g=k of the highest point PA in the contour line representation, and the point PA denotes a 1st representative color in each of the g-planes. Such process is executed with regard to the entire g-planes, i.e. to each value of k ranging from 0 to 31.
In a next step S612, lower appearance frequencies are cut off out of the above-described frequencies f (r, g, b) with regard to each g-plane. For example, on the g-plane where the g-axis address has a value k (k = 0 to 31), the appearance frequencies less than 1% of the maximum frequency at the point PA are cut off out of the aforementioned f (r, k, b) or fg=k (r, b) as fg=k (r, b) ~ 0.01 x max (fg=k (r~ b)) In this case, the aforementioned function is set as fg=k (r, b) = o As shown in Fig. 13B, this process corresponds to the removal of low level portions out of the all contour lines shown in Fig. 13A, and such process may be executed with regard to the entire pixels on each g-plane. Such removal or cut-off of lower appearance frequencies is necessary for averting any false contour or edge in the color image.
Subsequently in a step S613, a point farthest from the aforesaid point (e.g. PA in Fig. 13B) of the 1st representative color is selected as a 2nd represent-ative point on each g-plane. Relative to the significant region (e.g. island portion remaining in Fig. 13B) after the cut-off process executed in the step S612, the distance LB (r, b)g~ there are found the subregion coordinates (rB, bB)g where such maximum value is obtained. In Fig. 13B, the point having the longest distance from the point PA is selected as a point PB, which is determined to be a 2nd representative point on the g-plane.

In the significant region (island portion in Fig. 13B) left posterior to cut-off of the color appearance frequency distribution at a predetermined frequency with regard to the pixels corresponding to edge, contour or the like where the aforementioned deterioration is prone to occur, selection of any other representative points than the maximum frequency point PA is based on the condition that none of the display colors becomes conspicuously different from the original color and also that none of the points is extremely far from the representative color in the significant region.
For this reason, the point PB farthest from the 1st representative point PA is selected as a second repre-sentative point, and thereafter a point farthest from such two representative points is selected as a next representative point.
In a next step S614, the count parameter j is set to 3. This numerical value 3 corresponds to selec-tion of a 3rd representativé point in succession to the 1st representative point PA of the maximum appearance frequency and the 2nd representative point PB farthest from the point PA.
In a step S615, a point farthest from the fore-going representative points PA, PB and so forth is 1333~21 selected as a next representative point; and in a step S616, the count parameter j is incremented by 1 as (j = j + 1). And the routines of the steps S615 and S616 are executed repeatedly until the count parameter j is set to 6 in a next step S617 (or during the period of time in which the condition j < 5 is satisfied).
First in the step S615, all the distances from the points PA, PB and so forth are found in each sub-region of the significant region per g-plane after the frequency cut-off process, and the subregion having the m~x;mum distance is selected as a next representative point. In this case, the representative point obtained with j = 3 is selected as a point Pc in Fig. 13B, the point with j = 4 is selected as a point PD, and the point with j = 5 is selected as a point PE, respectively.
In selection of the five representative points, the lower appearance frequencies are previously removed, and the significant region posterior to such cut-off process is uniformly divided regardless of the appear-ance frequencies. In this embodiment, five represent-ative points are selected on each of the 32 g-planes, so that a to.tal of 160 points (representative colors) can be obtained in the color space.
A decision is made in a step S617 as to whether 1333~21 the count parameter k is equal to or greater than 6 (j 2 6), and when the result is yes, the process advances to a next step S618.
In the step S618, representative colors of primary red, green, blue data R*, G*, B* are allocated to the five representative points on each of the g-planes. Supposing that the 1st to 5th representative points are at the coordinates (r~, bj) (where j = 1 to 5) on the g-plane, the respective primary color data R*j, G*, B*j at the representative points can be obtained as follows:
R*j = (rj + 1/2) x (Rmax - Rmin + 1) / 25 + Rmin G* = (g + 1/2) x (Gmax - Gmin + 1) / 25 + Gmin B*j = (bj + 1/2) x (Bmax - Bmin + 1) / 25 + Bmin And such calculations are executed with regard to the entire 32 g-planes to consequently obtain 260 represent-ative colors corresponding to a total of 260 sets of the three primary color data (R*, G*, B*).
As described above, the five representative points on each g-plane are selected from the colors of the pixels where deterioration of the color image is prone to occur. Moreover the colors of the lower appear-ance frequencies are previously cut off, and the signifi-cant region posterior to such cut-off process is divided 133~421 uniformly regardless of the appearance frequencies, so that it becomes possible to eliminate any harmful influence of granular noise or false contour known here-tofore to cause some problems in the conventional method of division. And fast follow-up is also rendered pos-sible in combination with the above-described adaptive mapping by single color renewal.
The 160 representative colors thus selected are mapped by the aforementioned method to form the display data dd shown in Fig. 12A. With regard to any pixel where image deterioration is prone to occur or the pixel at the top of each horizontal line, another màpping is performed by the use of such 160 representative colors.
The pixels to be processed with the representative color mapping include, for example, those at the tops of individual horizontal lines and the pixels where the sum of the absolute values of the primary-color data differ-ences between mutually adjacent pixels become greater than a specific threshold value, as chosen in the steps S605 and S606 of Fig. 15. In addition, a modified method may also be carried out by first detecting, in the color space, the distance or error between the restored three primary color data obtained by the adaptive mapping of the data dd(r), dd(g), dd(b) and the 1333~21 three primary color data of the original color image;
then comparing, in the color space, such detected distance or error with that between the three primary color data obtained by the representative color mapping and the three primary color data of the original color image; and adopting, as the color display data, either of the primary color data to reduce the distance or error.
According to the color data compression method of this invention, as is apparent from the description given-hereinabove, adaptive mapping is performed, with regard to any portion of the color image having a large area with occurrence of gentle color change, so as to provide the present color display data on the basis of the difference between the mutually adjacent pixels, thereby achieving smooth display with satisfactory color reproducibility. Besides the above, there is previously detected any edge portion of the image or the top of each horizontal line in a frame where the difference between the restored data and the pre-compression original color data is rendered great in the adaptive mapping, and representative colors are selected on the basis of the pixel colors of such image portions. Then another mapping is performed by the use of such selected 1~3~421 representative colors, whereby adequate color repre-sentation can be accomplished even in any image portion where the color data difference becomes great in the aforementioned adaptive mapping, hence achieving remarkable improvements in the overall image quality.
It is to be understood that the present inven-tion is not limited to the above-described embodiments alone. And in accordance with any apparatus or system employed, some proper modifications may be contrived with respect to, for example, the number of bits of the digital primary color data or the compressed data per pixel, or the number of bits of the control code or the color saturation data mentioned in each of the embodi-ments.

Claims (3)

1. An apparatus for processing color display signals, comprising:
color data generating means for generating a plurality of color data signals, each consisting of a predetermined number of bits and representing a respective one of a plurality of color components of a corresponding pixel;
transmission data generating means for generating a series of transmission data signals based on said color data signals and each representing a respective pixel and consisting of no more than said predetermined number of bits, said transmission data generating means including a central processing unit programmed for comparing at least two of said color data signals, each representing a respective color component of a first pixel, with corresponding color data signals of a second pixel adjacent toand preceding said first pixel to produce respective differential values for each said comparison, said central processing unit being further programmed for selecting as the transmission data signal representing said first pixel, oneof said at least two color data signals of said first pixel having a maximum differential value with respect to said corresponding color data signals of saidsecond pixel, and for combining in each said transmission data signal identification data identifying the respective color component and code data corresponding to a value of the respective color component:
receiving means for receiving said transmission data signals for said pixels;
a plurality of memory means each for storing display data signals representing a respective one of said color components;
means responsive to said identification data in said selected one of the transmission data signals which is received for renewing said display data signals stored in a corresponding one of said memory means, the display data signals which are renewed corresponding with a color identified by said identification data and having a renewed value corresponding to said code data in the received transmission data signal; and means for supplying the display data signals stored in said memory means to a display means.

2. The apparatus according to claim 1, wherein said color data generating means is operative to generate said plurality of color data signals such that each of said color data signals represents a respective primary color.

3. The apparatus according to claim 1, wherein said central processing unit is programmed to generate said code data as a compressed form of said color data signals.