JPH01194757A - Color image processing system - Google Patents

Abstract

PURPOSE:To obtain an image with high picture quality and having an accurate color tone with simple constitution by providing a color tone rendering means to perform color tone rendering so that the color of a color image and that of an original image can coincide or neighbor with each other, and a color ghost correction means. CONSTITUTION:The color components of the original image are read by CCDs(4...6), respectively, and those output are inputted to a color tone rendering processing means 16 via A/D converters(7...9) and density conversion circuits (13...15). The color tone rendering processing means 16 performs the color tone rendering so that the color of the color image formed at an image output device 19 and that of the original image can coincide or neighbor with each other. The color image on which the color tone rendering is applied, after color ghost correction being applied at the color ghost correction means 17, is outputted to the image output device 19 via a multileveling formation circuit 18.

Description

[Detailed Description of the Invention] (Industrial Application Field) The present invention relates to a color image processing system used in copying machines, facsimile machines, etc., and more specifically, to a color image processing system that can accurately reproduce image information such as color tone. Related to color image processing systems.

(Background of the invention) Generally, color images include red (R) and green (G).

It is separated and read into three colors: blue (B), yellow (Y), and magenta (M), which correspond to the toner colors.

Usually, images are formed by converting into four colors: cyan (C) and black (K). That is, image data read as RGB positive values is generally converted into positive values for Y, M, and C, which correspond to the adhesion area ratio of each color toner.

Conventionally, as such a method, masking processing that takes into account the characteristics of toner has been considered effective. This is R, G, B
The conversion from Y, M, C, etc. is performed using a linear or quadratic or higher order matrix calculation.

In addition, in laser printers, etc., in order to change the density of the toner transferred to the transfer paper, by changing the pulse width of the multiplex laser, the amount of toner adhering to the drum can be changed in several stages. There is a way to express the tone. In addition, in color printers, toners of different colors (Y,
There is a method of overlapping M, C, K) on a drum and transferring them to transfer paper in one go.

By combining the above-mentioned methods, a multicolor hard copy can be produced relatively easily without using patterns such as dithering.

(Problem to be Solved by the Invention) However, when a plurality of color toners are superimposed on a drum, the developability is quite different between the color toner that is developed first and the toner that is superimposed and developed on top of it. come. For this reason, the relationship between the 1-inclusive laser and the transfer amount, therefore,
R, G, B to Y.

The relationship of conversion into M, C, and so on is complicated, and it may be difficult to approximate it using polynomials or the like.

By the way, when an image reading device scans and reads a color image, a phenomenon called color ghost is likely to occur due to misalignment of the scanning system and optical system. This is a phenomenon in which unnecessary chromatic colors appear particularly around the edges of the nine black characters, which adversely affects image quality. It is possible to predict the appearance of color ghosts to some extent based on the dot pattern of the output image, but this requires a means to determine whether the dots in the output image are chromatic or achromatic, and a large capacity memory to store the dot patterns. It is impossible to avoid increasing the scale and complexity of the circuit configuration.

The present invention has been made in view of the above-mentioned problems, and its purpose is to realize a color image processing system that can obtain print output with high image quality and accurate color tones with a simple circuit configuration. It's about doing.

(Means for Solving the Problems) The present invention to solve the above problems provides an image reading means for reading a document image for each of a plurality of color components and outputting digitized color image data for each color; and a color image forming means for forming a color image on a recording medium by sequentially overlapping images of a plurality of colors based on the colors of the color image formed by the color image forming means and The present invention is characterized in that it includes a color tone reproduction means that reproduces the color tone so that the color matches or is closest to the color of the original image, and a color ghost correction means that corrects the color ghost that occurs in the color image.

(Operation) The color reproduction means selects a color so that the color of the color image formed by the color image forming means and the color of the original image match or are closest to each other. The color ghost correction means corrects color ghosts that occur in color images.

(Example) Examples of the present invention will be described in detail below with reference to the drawings.

FIG. 1 is a block diagram showing the configuration of an embodiment of the present invention. In the figure, 1 is the original, 2 is a lens for forming an image of the original 1 on the CCD, and 3 is the image information of the original 1 in RGB.
A prism that separates into three primary colors, 4.5.6 is R
C0D converts image information decomposed into GB into electrical signals
17, 8.9 are A/D converters that convert RGB image information into digital signals, 10, 11.12 are shading correction circuits for shading correction, 13
．． 14 and 15 are density conversion circuits for data compression, and 16 is a color that converts RGB signals into Y (yellow), M (magenta), C (cyan), and K (black), which can be output by the image output device. 17 is a color ghost correction means. 1
78 to 17d are color ghost correctors, and 17e is a color ghost detector. On the 1st, a multi-value conversion circuit for converting image data into multi-value data for outputting to an image output device, 18
8 to 18d are multilevel ROM11819 are threshold 1iIROM
It is. Reference numeral 20 denotes an image output device for obtaining a hard copy of the original document 1 from multivalued image data.

The operation will be explained below. The image information of the original \a1 is separated into three color separated images by the prism 3. In this example, a color-separated image of red (R), a color-separated image of green (G), and a color-separated image of blue (B) are used.
It is separated into color separated images.

Therefore, the separation wavelength of RGB is 560 to 590 nm and 48 nm.
It is 0 to 500 nm. Each color separation image of RGB is C0D4
, 5.6, and each RGB image information is output.

The image information RGB is converted into a digital signal with a predetermined number of pits (for example, 8 pits) by a Δ/D converter 7.8.9, and at the same time, a shading correction circuit 10.11.
．． Shading correction is performed by 12. The shading-corrected digital image signal is converted into a 6-bit RG density signal and a 5-bit B density signal by a density conversion circuit. This density signal is converted by the color tone reproduction processing means 16 into color signals of the output colors (here, four colors of Y, M, and C) of the image output device. This tone reproduction processing means has an image processing ROM (not shown) that stores tone reproduction processing information. Among the color tone reproduction processing information, one unit is 8 pits of 6-bit color information and 2-bit color code, and there are 4 sets (for 4 colors of YMCK). This color tone reproduction processing information is used to maintain the same color tone as the original as much as possible.
If the color of the original cannot be output by the image output device 19, the color reproduction processing RO selects and outputs the color closest to the color of the original (hereinafter referred to as color reproduction or color tone reproduction).
It is stored in M. That is, by accessing the tone reproduction processing ROM using the density signal as an address, eight-pit color tone reproduction processing information can be obtained.

The obtained color tone reproduction processing information is subjected to color ghost correction by the color ghost correction means 17. That is,
The color tone reproduction processing means 16 separates the signal into a 2-bit color code and a 6-bit color signal, and the color hood is applied to a color ghost detection circuit 17e, and the color signal is applied to color ghost correctors 17a to 17d. A color ghost detection circuit 17e detects a color ghost, and color ghost correctors 17a to 17d correct the color ghost occurring in the color signal.

The color signal whose color ghost has been corrected is converted into a multi-value signal by a multi-value conversion circuit. This multi-value conversion circuit has a multi-value conversion ROM and a threshold value ROM, and the values of a threshold value matrix for multi-value dithering are recorded in the threshold value ROM. The values of this multi-value dither threshold matrix are supplied to the multi-value ROM in synchronization with the color signal by a synchronization signal (not shown). This multilevel ROM uses the color signal and the value of the threshold matrix as addresses, and outputs the determination result as a multilevel signal.

This multivalued signal is sent to an image output device 19 such as a printer and output as a dot pattern.

Hereinafter, the color tone reproduction by the color tone reproduction processing means 16 will be explained in detail.

In order to reproduce the same color tone as the original, the present invention selects a combination of input and output signals that minimizes the color difference based on a discrimination amount such as the actual color difference (ΔEa b is used here), and then selects the combination of input and output signals that minimizes the color difference. Store it in ROM in advance. A method of generating color tone reproduction processing information to be stored in the ROM will be described below.

■ Creation of color target/color measurement First, in order to examine the output characteristics of the image output device @19, a color target is created. In this embodiment, the image output device 19 is a YMC
It has the ability to output 11 degree steps for each of the four K values. Since this device superimposes the toners on each other, the colors that can be expressed with color toners are 44-256 colors. These colors are outputted to the image output device 19 to obtain a color target. The obtained color mark is placed on the document table of an image reader, and converted into a brightness signal of 8 bits each for RGB by scanning.

This RGB brightness signal is converted into CI EXYZ coordinates and saved as data.

(2) RGB-+XYZ Conversion Matrix Operation In order to convert the RGB signals of (2) into CIEXYZ coordinates, the characteristics of the image reader must be investigated. Therefore, approximately 20 pieces of colored paper are selected from the Munsell color standard and measured using a colorimeter to obtain values in the Cr EXYZ coordinate system of the colored paper. Next, by placing the colored paper on the document table of the image reader and scanning it, the image reader of the colored paper can perform the RGB
The brightness signal of is 1q.

In this way, the two values of colored paper obtained in ■ and ■ are as follows:
Since there is a linear relationship, the following equation holds true.

Here, the parameters a-, , i are obtained by approximation using the least squares method from the above two types of values. That is, a-i
By determining the parameters, the RGB signals from the image reader can be converted to XYZ, and the characteristics of the image reader can be investigated.

(2) Simulation printers that generate dot patterns are capable of displaying 256 colors with one dot as mentioned above, but they require the ability to display even more colors in order to reproduce colors.

In order to solve this problem, the present invention uses a four-level dither method. This uses three threshold matrices with a dot size of 4 x 4, and inputs integer values from 0 to 48,
It is possible to output a 4-level signal with a size of 4×4.

By the way, it takes a lot of effort to output this huge number of reproduced colors to a printer and measure all of them. Therefore, the generation of the dot pattern and the color measurement are all performed by simulation using calculations. A specific example will be explained.

The color signals before being seen in the multi-level conversion circuit are YMCK, 0 to 4 respectively.
Takes an integer value of 8. Incidentally, in this example, processing is performed to use the K (black) signal, that is, the black toner as much as possible. In color printing, Y (yellow) 9M
When magenta and cyan (magenta) inks overlap at the same location, it means black. Processing that replaces the black component with black ink and reduces the amount of other chromatic inks used is generally inking (UCA) and undercolor removal (UCR).
It is called. In the present invention, when the color signals YMC are all higher than the O level, a method is used in which the levels of the CMY signals are equally lowered so that one of the YMC becomes 0, and the black level of that dung is raised instead. .

The formula in Table V is as follows.

K+Pxmin (C, M, Y)='C-KX
S=C'M-KXS=M'Y-KXS=Y' Here, m1ni is a function that takes the minimum value among the numerical values within 0, and P is a parameter indicating the degree of toner replacement. S is a changeover switch between UCA and UCR, and 0
At 08:00, S=1. When in UCA, it becomes S-O. In this example, we set P=1.8=1, resulting in 100% UCR
was carried out. Also, in this example, ni' is YMC
In the above equation, -0 was set so that it is obtained only from the black component.

Therefore, the types of colors reproduced by the printer are narrowed down to 49 to the third power. However, it is safe to say that the number of reproduced colors is sufficient and does not affect color reproduction.

Next, a reproduced color dot pattern covering 49 cubes is generated. First, one YMC signal (0 to 48) is determined. For example, if Y-30, M=20, and C=10, then Y'=20. M' = 10.0' = 0. ni'−
converted to 10. These values of Y'M'C'K' are converted into a multivalued (0 to 3) matrix via the threshold matrix shown in FIG. 2, respectively. The threshold matrix consists of three, for example, the first matrix is 1 to 16.
, the second number is 17 to 32, and the third number is 33 to 48 arranged randomly. Here, I have just decided on Y.

If the values of M and C are the points in the upper left corner of the matrix, then
Since Y' is larger than 1 to 17 and smaller than 33, it becomes 2. Similarly, M' becomes 1, and C' and OSK' become 1. These four multi-value matrices of Y'M'C'K' are superimposed as shown in FIG. 3 to obtain a multi-value dot pattern.

Here, the value of C'M'Y'K' at the same position is ■
It corresponds to the color of the color target created in . For example, the value of C'M'Y'K' at the top left of the matrix is C'
=1. M'=2. If Y' = O, and Y' = 2, the level of cyan among the colors of the color mark ■ is at that position. The magenta level is 2. The corresponding color corresponds to a color in which the level of yellow is O1 and the level of black is 2.

Since the color of the color target has already been converted into the CfEXYZ value at the stage (■), a new matrix in which these values are arranged is created.

FIG. 4 is an explanatory diagram for explaining such a matrix. The color reproduced by a printer can be expressed as the average color of an area having a size of 4×4 dots as shown in FIG. Therefore, the CTEXYZ coordinate value of the color reproduced by the printer is
, Y, and Z, then X-(1/16)ΣXi 5l Y=(1/16)ΣYi ζ襦1 Z=(1/16)EZi i工1. This is accompanied by the condition that the size of the dots is constant during actual output. If the size of the dot is
If it differs depending on (■), the following formula may be used.

That is, However, 3i is the area indicated by the dot.

In this way, the average color of a small area, that is, the color reproduced by the printer, can be calculated at the simulation level. Similarly, CMY parameters are independently set to 0 to 48, and a total of 493 reproduced colors are calculated.

Now, with ■, the colors of the original read by the image reader can be converted to Cr EXYZ values, and with ■, all the reproduced colors when the printer uses a 4-value dither with a matrix size of 4 x 4 are converted to CIFXY
It is multiplied by 1η depending on the value of Z.

■Color Matching This section describes how to connect the data of ■ and ■ so as to maintain the same color tone as possible with the original, and obtain the relationship as color tone reproduction processing information. In this example, the original reading signal from the image reader is density-converted, and R=G=6bit
, B=5bit digital signal. The signal at this time is for all cases (R, G = O ~ 63
．． B=O~31). Then, each time, the following processing is performed.

The density-converted RGB signals are converted into XY2 signals by (1), and then converted into coordinates in a uniform color space. In this example, conversion to CIELAB uniform color space is performed, but CIELUV, LHC, etc. are also effective.

The conversion formula for the CrELAB uniform color space is as follows.

Including L = 116 (Y/Yo) -16a*=5
00 ((X/XO)'-(Y/Yo)')
b = 2oo ((Y/Yo)” - (Z/Z.)
°) Tree - jg Motoki Next, choose the color closest to the color represented by L, a, b of (from among the 49 cubed colors produced by the printer in ■). A discriminant amount to represent is necessary, which can be achieved by using Euclidean distance on a uniform color space.

The advantage of representing the signals to be compared on a uniform color space is that the uniform color space is designed so that the distance between two points in the uniform color space matches the human sense of color difference as much as possible. Because there is. Therefore, the distance closest to the color from the image reader is 1it (color difference, CIELA
In the B color space, this ΔEab) is the shortest, and calculation processing is performed to select it from among the colors reproduced by the printer, and the obtained relationship (RBG11 from the Image League side
The relationship between the color tone signal and the YMC signal representing the color reproduced by the printer) may be used as the color signal portion (6 bits) of the color tone reproduction processing information.

Note that the above method makes it possible to select and output the closest color even if the color of the document is not within the color gamut of the toner.

FIG. 5 is an explanatory diagram showing the state of this color tone reproduction. In this figure, the signal on the image reader side exists outside the reproduced color gamut of the printer, but the color with the smallest ΔEa*b# is selected as the reproduced color.

The smallest tree ΔEab means that the color is the most difficult to distinguish.

Problems in this case include the need for a large amount of computer processing time and the need for a human-sized memory for storing color tone reproduction processing information.

The former problem can be solved in several tens of minutes using a large-scale computer, and the latter problem can also be solved because the price of memory is decreasing.

(2) Color Ghost Processing Due to misalignment of the scanning system or optical system of the image reader, chromatic ghosts (hereinafter referred to as color ghosts) appear particularly around the edges of black areas. Among these color ghosts, the following are considered to be caused by image reading: (a) Pixel shift between each C0D4, 5.6 (b) Mismatch in lens magnification (c) Lens chromatic aberration (d) Superimposition of noise components, etc. It will be done.

An example of appearance of color ghosts is not shown in FIG. This figure shows color ghosts that appear after the 1lilfiJIL color tone reproduction process is applied to the black kanji character "sexuality".

Moreover, FIG. 7 is an enlarged view of a part of FIG. 6. As you can see from this example, the 7th color ghost
As shown in the figure, yellow and cyan appear at the edges of the black line.

It is clear that color ghosts appear differently in other color combinations.

The causes of such phenomena will be explained in detail for cases (a) to (c) above.

(a) Pixel misalignment of three CODs If the CODs are not precisely aligned, cyan, mapped, and yellow ghosts will appear at black edges during color tone reproduction processing.

Therefore, in order to prevent this, it is necessary to align the three CODs closely. Normally, it is necessary to perform alignment within one pixel, preferably within 1/4 pixel. In this example, this is achieved by aligning the three CODs on a jig and then fixing them with adhesive.

(b) When printing an original with a color that does not match the lens magnification, there are effects such as chromatic aberration of the lens. This is especially true because when the wavelength range of light is divided into two, green and red, the image formation position F on the edge side and the image formation position E /fi on the red copper are different, as shown in Fig. 8, for example. This is a phenomenon that is noticeable at high altitudes.

Depending on the lens, a shift of about one pixel may occur.

(C) Lens Chromatic Aberration Unless a design is made to improve lens chromatic aberration, the MTF values for lens chromatic aberration may vary greatly between red, green, and blue. This appears as a level difference in the CD output.

It is preferable to take into consideration when installing the COD 1) so that the difference in output signal level between red, green, and blue is as small as possible when the black line is imaged.

Although it is possible to reduce color ghosting to some extent by taking the measures described above, it is difficult to completely eliminate it in practice, considering the variation in lens performance during mass production and the variation in accuracy of COD mounting. .

For this reason, color ghost correction is also performed electrically using the color code after color tone reproduction processing.

Color ghost removal is performed using the color pattern method.

This is because the color boost color that appears is determined relative to the original color, such as a ghost from original black to chromatic color. When using the color pattern method, the color of the original image can be identified by examining the appearance (pattern) of the colors of the pixel of interest and surrounding pixels to determine the color of the pixel of interest.

As an example, FIG. 9 shows a pixel of interest, the surrounding color pattern, and the determination of the color of the pixel of interest determined at that time.

In the first (number 1) example, since both sides of the pixel of interest are white and black, the chromatic color of the pixel of interest is determined to be a color ghost appearing at the black edge. Therefore, in the first example, the pixel of interest is changed to black.

On the other hand, in the second (number 2) example, it is not determined that a color ghost has appeared, and the color of the pixel of interest is output as is.

Such processing is difficult to implement with a paralysis circuit, so in this example, it is stored in a ROM and used in LtJT (lookup table) format. One-dimensional and two-dimensional systems are considered as color patterns, but if the number of colors is N and the peripheral pixels including the pixel of interest are M, the number of color patterns is N8. Therefore, if a two-dimensional pattern is used, the number of M will suddenly increase, making it impractical.

In other words, in a two-dimensional pattern, although the number of peripheral pixels in each dimension (main scanning direction/sub-scanning direction) cannot be increased, the number of patterns increases.

FIG. 10 shows the relationship between size and number of color patterns.

In this example, the size is 1×7 in one dimension (that is, N
=4. M=7) color pattern is used, and color ghost removal is performed independently in the main scanning direction and the sub-scanning direction. At this time, since there is no difference in the appearance of color ghosts in the image between the main scanning direction and the sub-scanning direction, the same color pattern is used in the main scanning direction and the sub-scanning direction in this example.

The color pattern size is 1 x 7, but if the degree of color ghost appearance is small, 1 x 7 is selected.
It is also possible to use a smaller size color pattern, such as x5. Color ghosts of up to 1 pixel can be removed in a 1×5 color pattern, and up to 2 pixels in a 1×7 color pattern.

When a 1×7 color pattern is used, the color code is input as a ROM address. for example,
Color code: white 200, chromatic color 201, conversion color: 10
．． If black or gray: 11, the following color pattern is white white white black black surrounding pixels (target pixel (surrounding pixel color) color) color) color code pattern is white: white white black black black: Black: Black 00:00:
00:01:11:11:11, and the address is 07F. Also, the code of the converted color is stored at this address as shown in FIG. LLJT is executed using the above method.

In reality, a 1x7 pattern requires a 14-bit address line, and a bipolar ROM with a 14-bit address input and a 2-bit color code output is sufficient, but a high-speed ROM with this capacity is It is not widely available on the market and is expensive.

In the embodiment, the ROM is selected using the first pixel, and the LUT is performed using the codes of the remaining six pixels.

This color ghost correction is performed in the main scanning direction and sub-scanning direction of the read image, and when the final output code of the pixel of interest is converted color 10, the following color signal level conversion is performed.

When the recorded colors are Y, M, or C, the color signal level is set to O.

When the recorded color is K, add the color signal level (for example +16)
shall be.

In other words, by eliminating chromatic colors, the number of 11
This is compensated for by increasing the K color signal level. This can prevent the black line from becoming thinner.

FIG. 11 is a block diagram showing an example of the color ghost detector 17e. Color ghost processing is performed in the main scanning direction (
This is performed in the horizontal scanning direction) and in the sub-scanning direction (vertical scanning direction).

In this example, ghosts in the horizontal and vertical directions are removed using image data for 7 pixels in the horizontal direction and 7 lines in the vertical direction.

Color ghost processing applies only to color codes of image data.

Therefore, the color code read out from the color tone reproduction processing means 16 is first sequentially supplied to a shift register 20 having a 7-bit configuration and is converted into a ψ column for ghost correction in the main scanning direction. . This parallel color code data for seven pixels is supplied to a ROM 21 for removing ghosts in the horizontal direction, and ghost detection processing is performed for each pixel.

An example of how the ROM 21 is used is as described above.

The serially processed color code data is supplied to the line memory section 22.

The line memory section 22 is provided to remove vertical color ghosts using seven lines of image data. The gate group 23 has each line memory 2
Gate circuits 23a to 2 correspond to 2a to 22Q, respectively.
3 (I is provided.

The color code data for 7 line memories synchronized in the line memory section 22 is supplied to a ROM 24 for removing vertical ghosts in the next stage, where vertical color ghosts are detected and output.

FIG. 12 is a block diagram showing an example of the configuration of the color ghost correctors 178 to 17d. In this figure, 25
is YMCK based on YMCK signal and ghost detection signal
26 is a zero generator that outputs zero, 27 is an adder that adds +16, etc. to the K color signal, and 28 is a selector that switches in synchronization with the selector 27. When a color ghost is detected by the color ghost detector 17e, the selector is switched to a in the case of YMC based on the YMCK signal, and the YMCK signal is detected.
The MC color signal is set to zero, and in the case of K, the filler is switched to b, and addition (-t16) is performed to the K color signal. When color ghost does not occur, switch the selector to C,
Outputs YMCK color signals as they are. With such a configuration, color ghost can be corrected.

As described above, by printing out the signal that has undergone color tone reproduction and color ghost correction using an image output device,
It is possible to maintain high quality color images and maintain the same color tone as the original.

(Effects of the Invention) As described above in detail, according to the present invention, the color reproduction means is arranged so that the color of the color image outputted by the single-image output device matches or is closest to the color of the original image. By selecting a color and configuring the color ghost correction means to correct the color ghost that occurs in the color image,
With a simple circuit configuration, it is possible to realize a color image processing system that can obtain print output with high image quality and accurate color tone.

[Brief explanation of the drawing]

Fig. 1 is a block diagram showing the configuration of an embodiment of the present invention, Fig. 2 is an explanatory diagram of a threshold matrix, Fig. 3 is an explanatory diagram of the matrix, Fig. 4 is an explanatory diagram of the matrix, and Fig. 5 is a color illustration. An explanatory diagram showing the state of reproduction, Figs. 6 and 7 are explanatory diagrams of color ghosts, and Fig. 8 is an explanatory diagram of the occurrence of color ghosts.
9 and 10 are explanatory diagrams of color ghost correction, FIG. 11 is a block diagram showing the configuration of a color ghost detector, and FIG. 12 is a block diagram showing the configuration of a color ghost corrector. 1... Original 2... Lens 3... Prism 4.5.6... C0D7.8.9...
A/D converter 10.11.12...shading correction circuit 13.
14.15...Density conversion circuit 16...Tone reproduction processing means 17...Color ghost correction means 178-17d...Color ghost corrector 17Q...
・Color ghost detector 18...Multi-value conversion circuit 18a to 18d-...Multi-value conversion ROM 18e...
Threshold ROM 19... Image output device patent applicant
Konica Corporation Certification ceremony Patent attorney Fuji Ijima 1 person Figure 2 Figure 4 Figure 5 Figure L'* Figure 6 Figure 7 Figure 11 Color ghost detector in r

Claims (1)

[Claims] an image reading means for reading a document image for each of a plurality of color components and generating digitized color image data for each color; In a color image processing system having a color image forming means for forming a color image, a color tone is reproduced so that the color of the color image formed by the color image forming means and the color of the original image match or are closest to each other. A color image processing system comprising a reproduction means and a color ghost correction means for correcting color ghosts occurring in a color image.