JPS63114373A - Picture processing device - Google Patents

Abstract

PURPOSE:To attain a picture with high quality by storing picture element infor mation into a separate memory at each color and accessing the memories in common so as to eliminate color ghost by one original scanning. CONSTITUTION:A black separation ROM 91, a red separation ROM 92 and a blue separation ROM 93 store respectively density data for each color while a color separation map is being separated into black (achromatic) area, red area and blue area, and also store color code for color designation for each picture element. Luminance signal data outputted from latches 89, 90 and a color difference signal data are inputted respectively to the color separation ROMs 91-93 in common and the picture data outputted from the ROMs 91-93 are separated into color code for color designation and density data and they enter a color ghost elimination circuit 95 via synthesis circuits 94, 96.

Description

DETAILED DESCRIPTION OF THE INVENTION (Industrial Application Field) The present invention relates to a color image processing device capable of processing color images, and more specifically to a color image processing device capable of removing color ghosts and obtaining high quality color images. The present invention relates to an image processing device configured as described above.

(Background of the Invention) An image processing device is a device that reads a document image pixel by pixel using a photoelectric conversion element such as a COD, and performs image processing such as multi-value processing (including binarization processing; the same applies hereinafter). In particular, in the case of the color image 1115Ia, the above image processing is performed after the light image from the original image is separated into multiple colors (for example, red and cyan) using a color separation optical system.

First, color separation, which is the basic concept of the present invention, will be explained, and then color ghosts that occur during color image processing will be explained.

(1) Color separation Consider the case where color separation of black, blue (cyan), and red colors is performed. FIG. 4 is a diagram showing the spectral reflectance characteristics of each color chart of achromatic color, blue (blue), and red (red).

(a) shows the characteristics of an achromatic color, (b) shows the characteristics of blue, and (c) shows the characteristics of red. In both cases, the vertical axis is reflectance (%) and the horizontal axis is wavelength (rv). Further, fl is the characteristic of the sample, and fl is the characteristic of the reference density plate (here, the white density plate).

As is clear from the figure, each color has different spectral characteristics. Normalization is performed using the output value of a reference white density plate (one with a reflectance of 90% is used). Now, set the output values of the red channel and cyan channel to ■ Daikichi + VCW.
shall be. It is desired that the spectral characteristics of this reference density plate be flat, and this is schematically shown by f2 (dotted chain line in the figure).

The spectral characteristics of the guichroic mirror have a cutoff of 600
A beam with a characteristic of nm is used, and light with longer wavelengths than this enters the red channel, and light with shorter wavelengths enters the cyan channel. If the output values of the red channel and cyan channel of the color sample are V11' and VC', respectively, then the image data actually used V and ■c are as follows. '/Vc'11 =(
1) In this way, since Vc and V-sawa are each normalized using white having a reference angle of 11 degrees, 0≦V-sawa≦1.0 and 0≦Vc≦1.0. Therefore, by using O≦vR+vc≦2.0 as a measure to express the brightness signal, the reflectance (reflection density) of the original is determined to be at the 0-black paper level.

2.0 - Exists within the blank level. From the above, the luminance signal information can be expressed as VllVC...(2).

Here, since the spectral characteristics of the achromatic image are almost flat as shown in Figure 4 (a), VR' /VRW aVc' /Vc W -+V*
mV (...(3)

On the other hand, chromatic images exhibit unique characteristics, so red Vt
<' /V* * >V(H' /Vc VL→V large>Vc Cyan VR' /VRW <Vc ' /Vc W
→V scale<VC...(4) Therefore, it is also possible to adopt the VR/VC or Vc/V hole as an axis that simply represents the difference in color.

In this case, black type Vll /VCa 1. O or Vc/V
*a1.0 Cyan 0≦V* /Vc <1.0 or 1.0<V
C/VR≦o.

Red type 1.0<V*/Vc≦competition or 0≦VC/VR
< 1 , 0 - (5) Although it is separable, it is inconvenient to handle because it is not symmetrical with respect to the achromatic axis Z as shown in FIG. In FIG. 5, the vertical axis is luminance information (V+Vc), and the horizontal axis is color difference information (VR/VC). In this case, as is clear from equation (5), the color difference information (VR/VC) is diverged as ■. however,
It can be seen that color separation of black, cyan, and red colors is possible by finding some kind of ratio between VR and Vc in this way. Therefore, we will consider performing color separation based on the difference in the proportions of red and cyan components included in the entire luminance signal.

Since achromatic colors have flat spectral characteristics as shown in Figure 4 (a), VCaV
c holds true. Therefore, the luminance signal, that is, the total light [1(Vl
M of the vR9VC contained within the VC) is equivalent. In other words, the following formula holds true for black (achromatic) colors.

VC/(VR+VC)ao, 5 Vc/(VC+Vc)aQ, 5 =・(
6) - In universal chromatic colors, there is a difference in the collapse of VR and VC included in (VR+VC). In other words, for red series, 0.5<V trap/ (VR+VC) ≦1.00≦Vc
/ (VC+Vc)<Q, 5... (A) For cyan, 0≦VR/ (VFI +VC) <0.50.5<V
c/(Vlarge+Vc)≦1.0・(8) Therefore, VR/(VllVC) or Vc/ is used as the axis to express the color difference.
By adopting (V trap + Vc), it becomes possible to clearly separate black, cyan, and red colors. According to this method, since the ratio of v-contribution or Vc to (VllVc) is determined, it becomes symmetrical with respect to the achromatic color axis Z, as shown in FIG. 6, and does not diverge as described above.

As shown in the figure, using Vc / (VR + VC) on the color difference axis, red color O≦Vc / (VR + VC) <0.5 black color Vc / (VFI +VC> = Q, 5
Cyan 0.5<Vc/(VR+VC)≦1.0...(9
) and each color can be clearly separated. From the above, the color difference signal information is VR/(VIII +VC) or Vc/(VR+Vc) -< 1
0).

Next, the color gamuts of black, cyan, and red are determined based on the color separation method described above. An ideal achromatic writing instrument would be located on the achromatic axis Z in Figure 6, but in reality there are various black writing instruments and manuscripts, so Vc
/(VR+VC!>, the area has a width around 0.5.

Further, the area where the value is smaller than the (VR+Vc) value becomes blackish. Therefore, although there are some irregularities, the black (achromatic color) color gamut is basically T-shaped as shown in FIG. The shaded area in the figure is a black (achromatic) area.

The image that has been optically scanned and read is subjected to multivalue processing using predetermined equivalent values in order to be output and expressed. As explained above, it is possible to use (VR + VC), which represents the brightness level (reflectance) of the original, as a counterpoint for multi-value conversion, but it is better to correspond to the reflection density of the original and use this value. is easy to handle.

Therefore, for each color gamut of black, cyan, and red (V
It is necessary to clarify the correspondence between the value of R+Vc) and the 11 degrees of reflection of the original.

Therefore, the correspondence between the i-degree signal information (VR+VC) and the reflection density is performed for each color gamut of black, cyan, and red. When taking action, we separate the various pixels by color and compare the results plotted on the map with the dark side results. The am corresponding value is 4 pits (1
6 levels). If the 16 levels are expressed in hexadecimal notation from O to F, the corresponding density values for each color gamut will be as shown in FIG.

When a color separation map (ROM density correspondence values) corresponding to the color gamut shown in FIG. 7 is created based on the density correspondence values shown in FIG. 8, it will be as shown in FIG. The thick line in the figure indicates a temporary color gamut area.

FIGS. 10 and 11 are diagrams showing conventional configuration examples of image processing circuits based on the color separation map shown in FIG. 9. First, the circuit shown in FIG. 10 will be explained.

A light image separated into a red channel and a cyan channel by a color separation optical system (not shown) is sent to CCD1. It is charged and converted by the CCD 2, enters the amplifier section 3, and is charged to each amplifier 31 and 32.
is amplified by The amplified signals enter the A/D converter 4 and are converted into digital data by A/D converters 41 and 42, respectively.

Output VR of A/D converter 41 and A/D converter 4
The outputs Vc of 2 each enter the color separation information creating means 5 and are given as addresses to a VR+VC memory 51 and a Vc/(V scale+Vc) memory 52. The V+VC memory 51 outputs (vR+VC) data corresponding to the input address, and the Vc/(VR+Vc) memory 52 outputs Vc/(VR+VC) data corresponding to the input address. These data enter the color information storage means 6 and are given as addresses to a red/cyan memory 61 and a black memory 62, respectively.

Red Cyan Memory 61. The black memory 62 stores the density correspondence values for each color gamut of the color separation map shown in FIG. 9, and the density correspondence values corresponding to the input addresses are outputted and stored in the buffers 7.8, respectively. The density corresponding values stored in these buffers 7.8 are color select signals B.

A color selection circuit 1o receiving B and R selectively selects and outputs the color. In the case of the circuit shown in Figure 11,
The memory for storing density corresponding values is the red memory 63. Black memory 62. The other circuits are the same as in FIG. 10, except that the cyan memory 64 is divided for each color gamut, and buffer memories 7, 8, and 9 are added accordingly.

(2) Color ghost (a) Occurrence of color ghost When performing color separation of image data as described above, 2
Based on the outputs of the two CCD outputs R*Vc, calculate (VR+VC)...(11) as luminance signal information, and calculate Vc/(Vhole+Vc)...(12) as color difference signal information, and calculate these values. The color separation ROM as shown in FIG. 9 is accessed as an address. In this case, C
If one or both of the CD outputs causes an output fluctuation, the address in the color separation ROM will change.

For simplicity, it is assumed that both VFl and VC have changed by a small amount Δ (Δ<V hole, Vc). If the data after the change are VR' and VC', VR' and VO2 are respectively expressed by the following equations.

■Salt' -VR+Δ vc'-vc+Δ ...(13) Therefore, VR' +V(! '-VR+Vc +'2Δ
-(14)Vc' / (VR'+VC' ) -Vc (1+(Δ/Vc))X 1/ [VR+Vc (VR-Vc) Δ/ (V +Vc)
2)... (15) Therefore, the addresses of the color 11ROMs differ by the third term in equation (14) and the second term in equation (15).

From equations (14) and (15), the original image is red (7
) Rk: V* +Vc, Vc / (VR+VC
) changes in the increasing direction, and when it is cyan (blue), Vc/(VR+Vc) changes in the decreasing direction. FIG. 12 is a diagram showing the position change in the ROM table at this time.

Hereinafter, when the CCD output fluctuates from the normal value as in l:, the address in the color separation ROM changes. Therefore, if the color of the original image is near the boundary between each color when the CCD output value is normal, the color will change due to the output fluctuation (see FIG. 13). Unwanted colors generated in this way are called color ghosts.

(b) Actual cause of color ghost occurrence The actual cause of color ghost occurrence is pixel shift of 02 COD (
(COD installation misadjustment) ■ Improper adjustment of lens magnification ■ Lens chromatic aberration ■ Superposition of noise components other than the CODCCD output signal. Among them, the factor (■) comes down to the problem of mounting accuracy of multiple COD image sensors, and the pixel size of the COD image sensor (usually 7μ or 14μ square) when the same position is
This is an extremely difficult adjustment as it requires adjustment to an order of magnitude lower than that. It is also necessary to maintain this accuracy during use/transportation, which is an important issue in the design of color m-image equipment using multiple CODs. Each item will be explained below.

(b) In case of pixel shift between two CODs As shown in Figure 14, two CODs < cyan COD and red C
If the pixel of CO> is shifted by one pixel during installation, the black 2
When the main line is visualized as an Il image, the output of black and white becomes white, blue, black, red and white as shown in Figure 15 (a) (
(See Figure 13 represents white). In this way, a color different from the original image is output. At this time, it is actually difficult to accurately match one pixel, and pixel deviations such as 1/4 and 1/2 pixels occur.

Therefore, when the pixel shift m is small (it is considered to be within 1/4 pixel), the effect on output fluctuation is small and color ghosts are less likely to appear, but as the pixel shift becomes large, the Color ghosts appear, such as black and white black and white becomes black and white black red mortar → white blue black black red self. The same applies to the red and blue line patterns shown in (b) and (C). At this point, color ghosts appear due to the influence of output fluctuations, so color separation is 1! Colors located far from the color boundaries in the ffROM table are less likely to cause color ghosts.

Generally, as shown in FIG. 9, the width of the color separation region for black is narrow, so that color ghosts tend to appear for black. FIG. 16 shows a color ghost image when a pixel shift of 1/2 pixel occurs.

In the figure, area a is the original black area, area b is the red ghost area, and area C is the blue ghost area.) Lens magnification is not adjusted When capturing a color image, generally R, G, and B colors are used. It is desirable that the imaging positions for both are the same. On the other hand, due to lens design or processing, the imaging position is generally different for each color.

In terms of design, among the three R, G, and B (e line, e line.

Although the two imaging positions (of the F-line) can be made to coincide, the other will be different. Furthermore, from a processing perspective, even if the image formation positions match on the optical axis, they gradually become mismatched in areas where the image height is large. At this time, for example, if the mounting position of each COD is determined based on the focus, the magnification will be different and it will appear as if the effective pixel shift has become larger at the end of the COD. As a result of the above, color ghosts occur.

FIG. 17 shows this situation. Reflected light (light image) of original 71
After being focused by the lens 72, it is imaged on the C0D 73. In the figure, A is the normal imaging position, and B is the COD position when the subject is in focus. The broken line C is the imaging plane of the lens, LI is the normal length, and L2 is the length when the lens is focused on the subject. In this figure, the ccD that receives the cyan light image is located at position A.
If a COD that receives red light is installed at position B, the original image will be long in the COD that receives red light. Therefore, the color ghost becomes more noticeable toward the end of the COD.

This figure shows an example where the lens is slightly oversized,
On the other hand, when the image is slightly underexposed, the image becomes shorter and a similar effect occurs.

(c) Superposition of noise components When sudden noise or low-frequency noise is superimposed on one or both of the two outputs VR and VC, color ghosts (dots) occur when isolated in the former, and in the latter Color ghosts appear over a wide area. FIG. 18 is a diagram showing how color ghosts occur. (A) shows the state in which color ghosts occur when sudden noise is superimposed, and (B) shows the state in which color ghosts occur when low frequency noise is illuminated. In the case of (a), sudden noise occurs in the black solid area, and blue dotted color ghosts occur. In case (O), a color ghost occurs over a wide range of red (hatched area) in the solid black area.

(Problems to be Solved by the Invention) As described above, color ghosts occur for various reasons in the color image processing process. Various attempts have been made to take measures against such color ghosts. Most of the conventional color ghost countermeasures are applied to binary image data after color separation.
This method could not be applied to image data of about 8 bits. Furthermore, when the number of colors is two or more, for example, three to seven, it is difficult to determine which color is a ghost image for which color.

For example, the invention described in Japanese Patent Application Laid-open No. 59-128872 stores the combination of colors of surrounding pixels of a pixel of interest and the estimated color of the pixel of interest at that time, and stores the color of the pixel of interest and the estimated color information. The color of the pixel of interest is determined by performing logical operations on both, and color ghosts are removed. At this time, the image data is binarized and has two colors, and since it is necessary to make the black image clear, the logical operation is performed with priority given to black.

Let us consider the case where color ghosts of three colors are corrected using this method. FIG. 19 shows an example of the occurrence of color ghosts in this case, and there are three types (A) to (C). In both cases, the center is the original line width. Depending on the method of color separation or the magnitude of level fluctuations, color ghosts may appear only on one side of the line or may not appear at all (generally less likely to appear on red edges), but in general there is a possibility that color ghosts may appear. (Of course, this also depends on how the binarization threshold is determined). Therefore, if we go by the idea of black priority, (b) in Figure 19.

The color ghost in case (c) remains as it is, and this method has the disadvantage that the black ghost image increases. This situation is particularly likely to occur when the line width of the ghost color is close to the original line width.

In this method, the black and red table addresses are separate;
If you try to apply it to three or more colors, N is the number of colors,
When m is the number of peripheral pixels, (NXS+) address lines are required. For example, if N-4.degree. Furthermore, as the number of colors increases, the logic circuit becomes complex, and it becomes practically impossible to process one pixel with content of about 6 to 8 bits.

On the other hand, the invention described in JP-A-60-160774 clarifies this situation, and in the conventional method, one pixel is 6
When handling data in bits, even in the case of two colors, 6
It is stated that three inputs are required and it is not practical. The invention is based on the assumption that a two-color original is handled, and a circuit is provided to determine whether or not one of the two colors is a ghost, and if it is determined to be a ghost, it outputs data for the other color. There is. Therefore, not only binary data but also multi-value data can be handled for colors to which the ghost determination circuit is not attached.

However, in a general multicolor document, each color has its own shading, and the assumption that one color does not necessarily have halftones does not hold true. Furthermore, when the number of colors increases, the number of addresses in the color table increases as in the case of the invention described above, and although there is an advantage for one color, the same problem occurs. As explained above, when attempting to remove color ghosts by handling a large number of colors (three or more colors) as medium @tone image data, conventional methods are incomplete.

Furthermore, in conventional image processing methods, it is not possible to output color signals (density data) for each color at the same time, so it is impossible to remove color ghosts in one document scan.

The present invention has been made in view of these points;
The purpose is to realize an image processing apparatus that can reliably remove color ghosts and obtain high-quality images in one scan of a document.

(Means for Solving the Problems) The present invention solves the above-mentioned problems in an image processing device that captures an image and obtains a plurality of color signals. The present invention is characterized by having means for separating the pixel information into a color code for specifying a color and 111 data, and a means for performing color ghost processing based on the color hood for specifying a color. .

(Operation) According to the present invention, pixel information is stored in separate memories for each color gamut, and these memories are accessed in common based on the imaging signal when removing color ghosts.

(Contents of the Invention) (a) The present invention removes color ghosts when capturing a multicolor image (color separation). For this purpose, the following steps are taken.

(2) Conventionally, only density data was output during color separation, but a color code (color specification information) specifying a color is added above or below the density data during color separation.

Processing related to 0 color (for example, color ghost processing) is performed based on color food.

Processing regarding 0 image data is basically performed on density information.

(2) Color ghost processing creates a color pattern from the color codes of the pixel of interest and its surrounding pixels, and determines the color code of the pixel of interest based on this color pattern.

Next, color separation and color ghost processing used in the present invention will be explained.

(b) Color separation Conventional color separation is performed using a circuit as shown in FIG. 10.11. The contents of the black memory, red memory, and cyan memory at this time are, for example, in the case of the black memory, data is stored only in a T-shaped area as shown in FIG.
In the case of the figure, data for 1 pixel and 4 pits), and data for other areas are O. The same applies to red memory and cyan memory.

In the present invention, a color code (color designation information) representing each color is added to the density correspondence value (111 degree information) shown in FIG. Then, the m-degree information and color designation information are combined into image data for each pixel. For example, when considering red and blue as chromatic colors, the color code is white = (1,1)-1x2+1-3 black-(0,0)-0+O=0 red-(1,0)-1x2+O-2 blue =<0.1>-0+1-1. Then, the density corresponding value having the value D in FIG. 9 changes as follows (conventional) (invention) OD (black) D → 2D (red) 1D (blue). Also, for example, in the black memory, data is written in the form of OX (X-0-F) in a T-shape, but in areas other than the T-shape, values corresponding to the code of ゛30 and white are written. will be done.

FIG. 20 is a diagram showing the storage state of memory for each color gamut. (A) shows the storage state of the black memory, (B) shows the storage state of the red memory, and (C) shows the storage state of the blue memory. Among the data in the figure, X is density data (value corresponding to 11 degrees). Note that the order of the color designation information and the density information may be reversed.

(c) Color Ghost Processing In the present invention, the color appearance of the pixel of interest and its surrounding pixels,
The relationship between the actual colors of the pixel of interest is investigated, and the color of the pixel of interest is uniquely determined based on the appearance of a specific color (color pattern method).

For example, if we are looking at the color appearance of a pixel of interest (1 x 7 I) and the three surrounding pixels, when scanning is performed and the color appearance becomes white, white, blue, blue, black, red, white, the pixel of interest (4th The blue color of the pixel becomes black. Therefore, for example, if the data for blue at this time is 18", this processing will change the data to "08". Examples of color patterns include the combination shown in Figure 21. There is.

In general, if N colors (including white) are judged using M pixels, it is sufficient to have N'' color patterns. Therefore, as an example, if N-2 to 4. The number of color patterns shown in the figure is given.Since ghosts appear at N-3 or more, if the number of color patterns shown in FIG. 22 is prepared, color ghosts can be corrected. At this time, it is preferable that the number of M is small, but normally a value of M-5 to 7 is sufficient.
It is possible to correct an image with a larger amount of ghosts as the value of is larger, but normally it is possible to correct one pixel's worth of ghosts with M-5 and two pixels' worth of ghosts with M-7. Also, M
When the value is -3, lines may become thin during ghost correction, which is not very preferable. In the above process, it is necessary to set the density data to O for pixels whose color has changed from data other than white to white after ghost correction.

(Example) Hereinafter, an example of the present invention will be described in detail with reference to the drawings.

FIG. 1 is a block diagram showing an embodiment of the present invention. The numbers attached to the signal lines in the figure indicate the number of bits of the signal. The optical images separated into red and cyan by a color separation optical system (not shown) are photoelectrically converted by C0Ds 81 and 82, and then amplified to a predetermined level by amplifiers 83 and 84. The amplified image signals are each A/
It is converted into digital data by D converters 85 and 86.

The A/D converters 85 and 86 output the white base t! The full scale W4m is set in advance so that the rated output is achieved when m image signals from a 1m degree plate are input. Therefore, the A/D converter 85.8
The data VR and Vc outputted from 6 are normalized data.

VR and VC each store ri times signal data (V
R+VC) ROM87. Vc for storing color difference signal data
/ (VR+Vc) is commonly input to ROM88 as an address, and (VR+VC) data corresponding to the input address is input from (VR+VC) ROM87 to Vc / (
Vc/(Vq +Vc) data corresponding to the input address is output from the ROM 88, and is respectively entered into latches 89 and 90 and latched. The data latched in latches 89 and 90 is black molecule flROM9
1, Red separation ROM92 and pre-meeting 1! ! It is commonly input to the tROM93 as an address. In the explanation so far, CCC) 81.82 is C0D1.0CD in Figure 11.
2, amplifiers 83 and 84 are connected to amplifiers 31 and 32, and A/D
Converter 85.86 connects A/D converter 41.42 to (VR
+VC) ROM87 is VR+Vc)tMo')51,
Vc / (V scale + Vc) ROM88 is Vc / (VR
+VC) memory 52, black separation ROM 91 to black memory 62, red separation ROM 92 to red memory 63,
The pre-separated ROM 93 corresponds to the cyan memory 64, respectively.

Black separation ROM91. Red separation ROM92 and pre-separation ROM
Reference numeral 93 separates the color separation map explained with reference to FIG. 9 into a black (achromatic color) area, a red area, and a blue area, and independently stores density data (density corresponding values) for each color gamut. Further, in the present invention, in addition to the density data, the color code for specifying the color described above is also stored in pairs with the density data for each pixel. These ROM91~
93 stores dummy data (value O) at addresses other than its own area. And 8 lunches.

(VR+VC) data output from 90, Vc/ (
VR+Vc) Data is here Color separation ROM91~9
3 as an address, the image data stored at the address specified by the address is output. Color 1! Each of the ROMs 91 to 93 outputs image data according to the input address, but only one of the three outputs has any meaning.

Note that the ROM for luminance signal data, the ROM for color difference signal data, and the ROM for color separation may be constituted by one ROM.

In this way, the image data output from the mROMs 91 to 93 for each color is separated into a color code for specifying a color and density data, which are then processed separately. The color code is sent to a color ghost removal circuit 95 via a first synthesis circuit 94.
The density data enters the color ghost removal circuit 95 via the second synthesis circuit 96. The color ghost removal circuit 95 mainly performs a ghost removal operation of converting the color hood that should become a ghost into a color code of another color (a color code that matches the surrounding color) upon receiving the input color code and data. After performing the scanning in the scanning direction and the sub-scanning direction (described in detail), the color code and density data from which ghosts have been removed are output.

The color code and density data output from the color ghost removal circuit 97 are latched into a latch 96. Behind the latched data, the color code is input to the first decoder 98 to be black, red, and blue.

The full output is demodulated into four color specification signals. On the other hand, the separately given color select signals (B, B, R signals) are the second
It enters the decoder 99 and is also black.

It is demodulated into four color specification signals: red, blue, and full output.

For these color designation signals, designation signals of the same color are input to the OR gate 100, and these OR gates 100 output a black selection signal and a red selection signal, respectively.

Output as blue select signal and full output select signal.

The selection signal for each color output from the OR gate 100 enters the density signal separation circuit 101. On the other hand, the f1 degree data latched by the latch 97 also enters the density signal separation circuit 101, and the density signal separation circuit 101 receives these signals, selects and outputs the density data of each color (details will be described later).
. Then, the density data outputted from the density signal separation circuit 101 is compared with the threshold value outputted from the threshold value circuit 103 by a subsequent comparator 102, and converted into binary data. This binarized data will be output as image data. Note that the multi-value processing is not limited to binarization, but can also be ternary, quaternary, etc.

FIG. 2 is a diagram showing a detailed configuration example of the color ghost removal circuit 95. The circuit shown in the figure is designed to remove color ghosts in both the main scanning direction and the sub-scanning direction. The color code (2 bits) enters the first shift register 201 and is shifted by the data clock.

The shift register 201 outputs 2×7 (bit) parallel data obtained by delaying a 2-bit color code by a maximum of 7 bits. This parallel data enters a ghost removal ROM 202 that removes ghosts in the main scanning direction. The ghost removal ROM 202 performs ghost removal processing using the color pattern method described above. That is, for a 1×7 color pattern, the color code of the pixel of interest is converted into a color code that does not cause color ghosts based on the color of the center pixel of interest and the colors of the three pixels before and after it. The color code that has been subjected to ghost removal processing in the main scanning direction in this manner is latched by the latch 203. III data (4 bits) enters the second 5-bit shift register 204, is converted to serial data by the data clock, and is sent to the latch 20.
It is latched by 3.

On the other hand, the previous number of one line, which is separately output, is counted by a 7-line counter 205, and a signal specifying a 1-line memory, which will be described later, is output from the 7-line counter 205. Then, the output of the 7 line counter 205 (3 bits)
enters the decoder 206 and is decoded. Decoder 2
Of the seven 1-line memory selection signals (hereinafter referred to as memory select signals) output from 06, one set is input to gates 01-G7 and 1-line memory Ml-M7, and the other set is input to gates G11-G7. Gt 7, G2 t~G
27, and other sets are input to the data selector 207.

Color code and I latched by latch 203
It is controlled by a memory select signal from the ff data decoder 206. The memory select signal of the decoder 206 sequentially selects the gates G+ to G7 line by line, and puts the one line memory connected to the selected gate into a write state. Therefore, 1 line memory M! ~M7 are each in a write state for one line out of seven lines, and the other lines output the written data. Similarly, the gates Gll to Gt7 are connected to the color code input/output lines of the 1-line memory, and operate to close the gates only when the 1-line memory is in the writing state. Gates G21 to G27 are connected to the ay1 data input/output line of the 1-line memory, and the 111th gate of the pixel of interest
1! Opens the gate to output data.

Therefore, the signals output from these memory groups are color code data (2
bit X 7-14bit) and the density data of the pixel of interest corresponding thereto. Therefore, the data selector 207 serves to rearrange the color code data whose sub-scanning order changes for each line based on the output of the decoder 206.

The write operation of 1-line memory Ml-MY is performed by Ml, M2.
．． ...+Me+MY, Mt,... are performed in this order. If the 1-line memory M1 is currently in the writing state while the document is being read, the 1-line memory M1
2 is the signal 6 lines before the run written in the 1-line memory M1, similarly, the 1-line memory M3 has 5 lines, the 1-line memory M4 has 4 lines, and the 1-line memory M5
The previous signal is stored in 3 lines, 2 lines in 1-line memory M6, and 1 line in 1-line memory M7, and read out by an address (13 bits).

The density data of the pixel of interest at this time is the data stored in the one-line memory M4.

The color code (14 pits) for each line output from the data selector 207 is entered as an address in the ghost removal ROM 208 in the sub-scanning direction, and color ghost correction of the target pixel in the sub-scanning direction is then performed.

As a result, the color code output from the ghost removal ROM 208 has been subjected to color ghost correction in both the main scanning direction and the sub-scanning direction. The color code for which this color ghost correction was performed is Gate G.
21-G It is once latched in the latch 209 together with the density data output from 27, and then output. Seven lines of color hood and density data are stored in the one-line memories M1 to M7, and read out and processed as needed. Since an actual image has hundreds of lines worth of text, it cannot be processed using the number of line memories shown in the figure. Therefore, an overlap operation is performed in which the last stored one-line memory is replaced with new density data.

FIG. 3 is a diagram showing a detailed configuration example of the 'fA degree signal separation circuit 101. The density data Do to D3 output from the latch 209 (see FIG. 2) are black and blue.

Red, data selector DS1 provided corresponding to all outputs
- Commonly input to DS4. On the other hand, the color hood and color select signals are input to a decoder 98.99 and decoded. Here, the correspondence relationship between the color code, color select signal, and color is as follows.

(a) Color code 00 black 01 blue 10 red 11 white (b) Color selection signal 00 black 01 blue 10 red 11 The color signals output from all output decoders 98 and 99 are signals output from common output terminals. Comrades (for example, the 0 output terminal of decoder 98 and the output terminal of decoder 9900)
is included in ORGATE Gst~G33. The logical OR outputs of these agates Gst to G33 are respectively input to the data selectors DS+ to DS3 in order as enable signals. However, all output terminals of the decoder 99 are directly connected to the data selector DS4.

In a circuit configured in this way, when a color is designated by either a color code or an external color select signal, the corresponding color select signal of either decoder 98 or 99 becomes "O", and the OR gate is activated. Predetermined data selectors DS1 to DS4 are enabled via G31 to Gss (however, DS4 is directly enabled by the output of the decoder 99). Density data is output from the enabled data selector and applied to the subsequent comparator 102 (see FIG. 1).

In addition, when all data selector DSs to DS4 outputs are high impedance (outputs are high impedance when not enabled), the output result is fixed to 0'' (white) for each signal line by the pull-down resistor 210. It is now possible to do so.

(Effects of the Invention) As described in detail above, according to the present invention, in order to obtain color separation data for each color, the l1 degree data for each color gamut based on the color separation map is combined with the color code and separately By storing the information in the memory, these color separation memories can be accessed in common during image capturing. Therefore, it is possible to realize an image processing apparatus that can reliably remove color ghosts and obtain high-quality images by scanning a document once.

[Brief explanation of the drawing]

FIG. 1 is a configuration block diagram showing one embodiment of the present invention, and FIG.
3 is a diagram showing a detailed configuration example of a color ghost removal circuit, FIG. 3 is a diagram showing a detailed configuration example of a density signal separation circuit, FIG. 4 is a diagram showing reflectance of each color chart, and FIGS. 5 to 7
The figure shows an example of a color separation map for each color gamut, Figure 8 shows a temperature correspondence value for each color gamut, Figure 9 shows an example of a color separation map, and Figures 10 and 11 are conventional Figure 12 is a diagram showing an example of the circuit of the device, Figure 12 is a diagram showing changes in the address position in the ROM table, Figure 13 is an explanatory diagram of color ghost generation, and Figure 14 is a diagram showing an example of a two-line pattern. The figure shows color data after color minutes + m, Figure 16 does not show color ghost images, Figure 17 is an illustration of pixel shift when lens magnification is not adjusted, Figures 18 and 19 20 is a diagram showing an example of a color ghost image, FIG. 20 is a diagram showing an example of arrangement of density data for each color gamut, FIG. 21 is a diagram showing an example of a color pattern, and FIG. 22 is a diagram showing the number of color patterns. . 1.2.73.81.82...C0D3...Amplification section 4...A/D conversion section 5...Color separation information creation means 6...Color information storage means 7.8.9. ...Buffer 10...Color select circuit 31.32.83.84...Amplifier 41.42.85.86...A/D converter 51...
VR+Vc memory 52-Vc/(VR+VC) memory 61...Red/cyan memory 62...Black memory 63...Red memory 64...Cyan memory 72...Lens 87-(V large+Vc) ROM 88・"Vc / (V scale + Vc) ROM89.90
．． 97.203.209...Latch 91...Black separation ROM 92...Red separation ROM 93...Nowaki IR0M 94.96...Composition circuit 95...Color ghost removal circuit 98.99, 206... Decoder 100... OR gate 101... Concentration signal separation circuit 102... Comparator 103... Threshold circuit 201.204... Shift register 202. 208-: f-st removal ROM 205... 7
Line counter 207--Data selector 210...Pull-down resistor M1-M7...1 line memory Gt-G7 +Gt+'Gtt, Gzt-
Gzv*G31~G33...Gate patent applicant Konishiroku Photo Industry Co., Ltd. Representative Patent attorney Fuji Ijima Tribute to 16 people from outside the country Corner diagram 420 (A) (Black memory) (B) (C) (Blue memory)

Claims (1)

[Claims] In an image processing device that captures an image and obtains a plurality of color signals, one pixel information is given as a color specification color code and density data, and this pixel information is separated into a color specification color code and density data. and means for performing color ghost processing based on these color specification color codes.