JPH01126078A - Color image processing method - Google Patents

Abstract

PURPOSE:To correct the position of a picture element without deteriorating the sharpness of an image by executing the shifting processing of the image and variable power processing at the same time. CONSTITUTION:The image data color-separated to R, G and B by an input system 1 is sent to an output system 9 through a shading correction circuit 2, a position dislocation correction circuit 3, an MTF correction circuit 4, a variable power circuit 5, a (gamma) correction circuit 6, a color correction.UCR circuit 7 and a gradation correction circuit 8. The position dislocation correction circuit 3 obtains the image data when the other color is shifted only prescribed shift quantity for a reference color by referring the data of the plural picture elements and this shifting processing of the image and the variable power processing by the variable power circuit 5 are executed at the same time.

Description

TECHNICAL FIELD The present invention relates to a color image processing method, and more particularly to a color image processing method suitable for a digital color image reading device and the like.

BACKGROUND OF THE INVENTION In manuscripts containing both pictures and text, the text is often black, even if it is a full-color manuscript.

To reproduce black on a full-color copier, use Yellow.

Three color materials, Magenta and Cyan (hereinafter referred to as rY, M, and CJ), are superimposed. However, Y, M,
Even if the three C colors are layered, if Y, M, and C are not perfectly balanced, some color components will remain. Also, Y
, M, and C, and if there is a positional shift, the image quality of images that require high resolution, such as characters, will deteriorate significantly. As a countermeasure for this problem, in digital color copying machines, it is possible to apply UCR (undercolor removal) in which the overlapping portions of the three colors Y, M, and C are replaced with black (BK).

On the other hand, as an example of a color image input method, a method is known in which a dichroic prism disclosed in Japanese Patent Laid-Open No. 187180/1980 is used to form an image on three CCDs to obtain a color signal. . In this method, it is important to completely eliminate chromatic aberration in the imaging optical system; one point on the document will not be accurately focused on a predetermined position I on each CCD, and therefore the color balance of the read pixel signal will be affected. You'll go crazy. In addition, the COD position adjustment is done using national electricity, and there are many parts.
There are also disadvantages such as high cost.

In another example, in Japanese Patent Application Laid-Open No. 61-61561, Red, Green, and B 1u are used as elements of a photoelectric conversion array.
e (hereinafter referred to as “R.

There is a system in which filters (referred to as G and BJ) are arranged regularly to solve the problem of position adjustment and number of parts. This method has another problem: since the R, G, , and B photoreceptors are shifted from each other, it is impossible to read data at the same point on the original image, and color correction processing and the above-mentioned OCR processing cannot be performed accurately. This was problematic.

Regarding this problem, Mizude Manjin first applied for
There is a method proposed in No. 94712 "Color image processing method". This method uses R,
Image data at a predetermined position is obtained by linearly interpolating the dot position shift between G and 8 using data of two adjacent pixels. With this method, pixel positions can be corrected, but linear interpolation means weighting is taking an average, and as a digital filter it has low-pass characteristics, so high frequency components of one image are suppressed. , there was a problem that the sharpness of characters, line drawings, etc. deteriorated.

This will be explained in detail below. Figure 20 shows G
, B, and R color separation filters are regularly arranged in a one-dimensional color image sensor (line sensor). When subdivided into G, It is composed of three minute pixels with B and R color separation filters. These minute pixels are arranged continuously in the main scanning direction.

Therefore, the sampling point of the image is shifted by 173 pixels between G-B and B-R, and by 273 pixels between G-R. Let us consider the influence of this positional shift on the UCR processing. Specifically, the sampling density is 16d
ot/mm, and as an image, the line thickness is about 100 μm, especially for black characters where UCR processing is a problem.
It will look like Figure 1.

Here, to simplify the explanation, consider a case where the output value of the image is linear in reflectance, and the coloring material, paper, etc. are ideal. 21st
In the figure, the black line spans pixels n and n+1. In an ideal sensor with no positional deviation between each color, 1 pixel n, n+1
The output value (reflectance %) is 5 for each of G, B, and R.
0% and 17%. When reproducing this using M, Y, and C coloring materials, which are complementary colors of G, B, and H, it is sufficient to use coloring materials having an area ratio of 50% and 83%, respectively.

Furthermore, if U CR 100% processing is performed, the minimum values of M, Y, and C can be replaced with BK, so in the end, only BK needs to be used at 50% and 83%. However, in reality, because there is a positional shift between R, G, and , C become as shown in Fig. 23, and even if U CR 100% processing is performed, 1M, Y, C
The amount of color material required for BK is as shown in FIG. 24, and it can be seen that BKI color cannot be reproduced.

In the above-mentioned Japanese Patent Application No. 194712/1988 entitled "Color Image Processing Method", linear interpolation is performed using data between two adjacent pixels in order to correct positional deviation between each color. Considering the sensor position of B color as the center, G and R colors have a positional shift of -1/3 pixel and 173 pixels, respectively. If this is corrected by linear interpolation, the linear formula % formula % Here, Gn, R, 1, etc. are the n-th of G color, the n-th of R color
Represents the output value of the first pixel, G n' + B n
'+Rn' represents a correction value.

The Gn and Rn correction processes described above are [0, 2/3], respectively.
．． 1/3], [1/3.2/3. Equivalent to one-dimensional digital filtering with coefficients of ◯]. The spatial frequency characteristics (MTF) of these filters can be found by performing discrete Fourier transform (DFT).

When the filter characteristic is expressed by f (x), its DFT is: x=1 U: Spatial frequency N: 1/ΔX ΔX: Sampling pitch m: Filter size MTF(u)=IF(u)I ------( a).

Figure 25 shows the MTF characteristics of the digital filter when F(x) = [1/3.2/3.0] m = 3 Δx = 62.5 μm (16 dots/am) (f(x) = [0
, 2/3.1/3] have the same characteristics).

As is clear from the figure, this filter has poor transmission of high-frequency components and has low-pass characteristics.

Therefore, by correcting the positional deviation using this filter, the high frequency components of the G and R color images are degraded. That is, the G and R images become blurred, and the sharpness of characters, line drawings, etc. decreases.

Purpose The present invention has been made in view of the above circumstances, and its purpose is to solve the above-mentioned problems in the conventional technology and to correct the position of one pixel without deteriorating the sharpness of the image. An object of the present invention is to provide a color image processing method that obtains a color image.

Structure The above object of the present invention is to provide a color image processing device having an image reading means using a one-dimensional color image sensor in which color separation filters are regularly arranged, in which another color is shifted by a predetermined shift amount with respect to a reference color. means for obtaining image data by interpolating with reference to data of a plurality of pixels, and means for performing scaling processing at least in the main scanning direction, and simultaneously performing image shifting processing and scaling processing. This is achieved by a color image processing method characterized in that it is configured to perform.

Following a detailed explanation of the present invention, based on examples,
The configuration will be explained in more detail.

According to the sampling theorem, the one-dimensional signal g(t) whose band is limited below the frequency W has an interval of T=1/(2W).

In other words, if sampling is performed at a sampling frequency of f = 2W, 1
The following equation allows the original signal to be completely restored.

T=1/(2W) According to the color image processing method according to the present invention, the principle of the above equation (4) is applied to the sampled image to obtain interpolated data, and positional deviation correction is performed.

Here, a method for calculating interpolated data will be explained. Third
The figure shows the relationship between the position of the sampling pixel, the position at which interpolated data is obtained, and other parameters. Using the parameters shown in FIG. 3, interpolated data ○(X) can be obtained by the following equation.

...(5) Here, h(x) is an interpolation function, h (x) = sinc (x) 1nnx x to ■, but from the graph of h(x) shown in Figure 4, As shown, when lxl becomes large.

h(x)→O. Therefore, in practice, 1 may be in the range of about -2 to 2.

By the way, the interpolation data can be any one of R, G, and B.
It is sufficient to calculate the remaining two colors using the color as a reference.Here, a case will be described in which positional deviation correction is performed for G and R using B as a reference. However, the positional relationship of R, G, and B is as shown in FIG. 20. That is, R
The data is the pixel of interest 1i! Interpolated data at a position shifted 173 pixels to the left from xo will be used.

Here, if xo is the origin (x,=O), the position from which interpolated data is obtained is x=--p. By using the above equation (5), the interpolated data becomes as follows.

FIG. 14 shows the values of each coefficient when using five pixel data of 1=-2, -1, 0, 1, and 2. One-dimensional filter [10,1676°0
．． 4191,0.8383. -0.2095,0.11
97], an image shifted by 173 pixels can be obtained. In other words, it is possible to convert the image into an image having the same phase as the B color.

When the MTF characteristic of this correction filter is determined using the above-mentioned equations (2) and (3), it becomes as shown in FIG. 5(a). It can be seen that flat characteristics without deterioration can be obtained up to the high frequency range. When implementing this filter in hardware, it is desirable that the coefficients be rational numbers that are as simple as possible. 14 shows an example of approximation to a rational number. The MTF characteristics of the filter using this coefficient are shown in Figure 5 (b
).

One thing to keep in mind when approximating coefficients to rational numbers is the ratio between coefficients. Let C be the coefficient for pixel i.

Therefore, although some intermediate calculations are omitted, it becomes , and l ill jl is small, that is, C□r C
For large ranges of j, the ratio of the coefficients becomes a simple integer ratio.

This makes it possible to ensure the accuracy of the position (coordinates) for which interpolated data is obtained.

In the above explanation, interpolation of R color data has been described, but for G color, the filter coefficient can be calculated using equation (6) with x=10-p. Also, G, B
, R, using the R color as a reference can be similarly calculated using equation (6) by setting x=+−p. By the method described above, it is possible to obtain a correction filter with less MTF deterioration.

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

FIG. 1(a) is a block diagram of a digital color copying system which is an embodiment of the present invention. In this system, image data that has been color-separated into R, G, and B by an input system 1 and quantized is corrected by a shading correction circuit 2 for unevenness in sensitivity of a light source and an image sensor.

Below, positional deviation correction circuit 3, MTF correction circuit 4, magnification changing circuit 5. The image data, which has been binarized (or multi-valued from 3 to 8 values), is passed through the γ correction circuit 62, the color correction/UCR circuit 7, and the gradation processing circuit 8, and is sent to the output system 9, where a copy image is obtained. It will be done.

Here, the MTF correction circuit 4 corrects image blur caused by the input system, and in a digital copying system, a digital filter having coefficients as shown in FIGS. 2(a) to 2(d) is used. The value of the coefficient should be determined based on the MTF characteristics of the actual input system, but the size of the filter is not limited to 3×3, and a filter with a size of 5×3, 5×5, etc. may be used. In addition, as shown in FIGS. 3(e) and (f), correction can be performed separately in the main scanning direction and the sub-scanning direction.

Further, in FIG. 1(b), the arrangement of the variable magnification circuit 5 and the MTF correction circuit 4 are exchanged, and the processing order of the variable magnification process and the MTF correction process is reversed.

In digital copying systems, it is relatively easy to obtain a wide range of magnification ratios, not only when using an equal-magnification optical system for the input system but also when using a reduction optical system. A method of varying the magnification is used.

The electrical scaling method converts a pixel data string sampled at the same magnification into a data string with a pixel density corresponding to the scaling factor. For example, when 100 pixel data is sampled at the same magnification, if you want to enlarge it by 120%,
Based on the pixel data, calculate 120 pixel data and
In the case of reduction by %, an operation is performed to obtain data for 70 pixels.

As an example of the above-mentioned electric magnification device, the present applicant has previously described
Patent Application No. 61-100503, No. 61-100505, No. 61-100506, No. 61-101721,
There are devices proposed in No. 61-104014 and No. 62-3262.

Among electrical scaling methods, those that use a Sinc function as an interpolation function can obtain scaling data with high accuracy.

Figure 6 shows the sampling position of the same size data (X-factor) and 1
Sampling position 11 (yi) when enlarged by 33%, sampling position when reduced by 80%! It shows the relationship (zi).

In this scaling method, image data at the positions indicated by yi and zl are obtained according to the above-mentioned equation (5) using data at the position indicated by xi.

By changing the density of the sampling points of the scaled image represented by yi+Zi, it is possible to scale the image at any scale ratio.

In this embodiment, by shifting the new sampling position after the above-mentioned scaling by a predetermined amount for each color, R, G, B
This method corrects the positional deviation between pixels at the same time as the scaling process. In other words, if we consider the relationship between the sampling positions of the three colors in the case of Figure 20 mentioned above, the pixel data at the position shifted 1x3 pixels to the left for R color and 173 pixels to the right for G color with respect to B color. It is obtained by interpolating.

1(a) and 1(b) show a digital copying system having a variable magnification processing section, but in this embodiment, positional deviation correction is taken into account during the variable magnification processing, so the system shown in FIG. 1(c) It is also possible to combine the image processing sections into one as shown in FIG.

By the way, the positional deviation correction circuit according to the present invention is a one-dimensional filter, and can be combined with a filter for MTF correction in the next stage to form one filter. FIG. 2(h) shows a combination of the MTF correction filter (a) and the positional deviation correction filter (g). In addition, (i) is a filter for MTF correction in the main scanning direction (
e) and a positional deviation correction filter (g).

In Fig. 2 (h), the size of 7 x 3 is obtained by combining filters of size 3 x 3 and 5 x 1, and in Fig. 2 (i),
By combining the 3X1 and 5X1 filters, a 7X1 size filter is obtained.

Here, the coefficients on the outside of the filter may be omitted considering that they are smaller than the coefficients near the center. FIG. 2(j) is a 5×1 filter obtained by ignoring the two outer elements of FIG. 2(i). Regarding (h), a similar simplification is possible. By combining filters into one in this way, hardware can be simplified and calculation accuracy can be improved.

As described above, the positional deviation correction process can be combined with the MTF correction process or the variable magnification process.

The combination with the MTF correction process is to superimpose the function for MTF correction on the interpolation function shown in equation (5), and the combination with the scaling process is to superimpose the interpolation data in equation (5). This is to shift the sampling position to be determined.

Therefore, if these two sets of combinations are performed at the same time, it is possible to combine the three processing units of positional deviation correction, MTF correction, and variable magnification into one. 6 Figure 1 (d) shows the positional deviation correction,
An example of the configuration of a copying system having a processing section that combines three processing sections for MTF correction and variable magnification is shown. In this case, the MTF correction circuit 4 in the subsequent stage performs MTF correction in the sub-scanning direction. Although it is possible to combine the MTF correction in the sub-scanning direction into one, there is an advantage that the correction calculation unit is simpler if it is separated by one.

A specific configuration example will be described below.

FIG. 7(a) shows an example of the circuit configuration of the positional deviation correction filter shown in FIG. 2(g). By receiving the image data output from the input system with the five stages of latches, it is possible to simultaneously refer to the data of five pixels in the main scanning direction. In this example, the image data is multiplied by the coefficient of the positional deviation correction filter shown in FIG. 2(g) using a ROM instead of a multiplier and using a table reference method.

Further, So and S-, , Sl and S-2 are made into pairs, and the first stage addition is also performed at the same time. That is, S
The sum of the products of the respective image data and the corresponding coefficients is stored in the memory address addressed by the image data of o and S-□ or S and S-2.

The intermediate results of the operation output from the ROM are added by the next adder, and the operation ends.

A filter that combines positional deviation correction and MTF correction in the main scanning direction (see Fig. 2 (i) and (j)) is shown in Fig. 7 (a).
) can be realized by expanding the circuit. In other words, you only need to add the required number of latches, multipliers and adders in the subsequent stage, and the difference in coefficients is as shown in Figure 7(a). All you have to do is change it.

FIG. 7(b) shows a configuration example of the MTF correction circuit in the sub-scanning direction. By sequentially updating and storing image data in the line memory for two lines, it is possible to refer to the data of three corresponding pixels of three consecutive lines at the same time.

Here, the MTF correction filter of FIG. 2(f) will be explained. Data of two surrounding pixels corresponding to a coefficient of -1 are added by an adder. The next stage ROM is for calculating the data of the center pixel and two surrounding pixels;
The operation of subtracting the output value from the adder from the value three times the value of the center pixel is performed using a table reference method.

In this way, when MTF correction is performed separately in the main scanning direction and the sub-scanning direction, either direction may be performed first. In other words, either of the circuits in FIGS. 7(a) and 7(b) may be placed first.

Next, the scaling process will be explained.

The scaling process is performed, for example, by a circuit configured as shown in FIG. The circuit shown in Figure 8 consists of a function to determine the position of a new sampling point after scaling, and a function to calculate new data from the distance between the new sampling point and the old sampling point and the old data. .

First, in order to extract peripheral data at once when determining a new sampling point and performing calculations in the future, the data synthesis unit uses a correction method to group data into 6-pixel units using the interpolation method using 6 peripheral pixels. . For example, in Figure 3,
When the new sampling point X is between xo and x-1, the arithmetic unit calculates S-3 and S-.

, S-4, So, and S□+Sz are taken out at once.

A specific method is to input data (DATAI) sequentially input in synchronization with a data clock to a data clock (DCL).
This can be implemented by latching at K). If there are 6 pixels, this can be realized using 5 stages of latches.

Next is the line memory section. This is a memory that stores groups of 4 to 8 pixels for the number of pixels in one line. It has a two-stage configuration for input and output, and when one is input, the other is The configuration is such that the input and output are reversed when one output line ends.

When input, the address of this line memory is
The address obtained by counting up the counter in synchronization with K is used as is, but this address is changed when outputting. The address at the time of output corresponds to the position determination function of the new sampling point.

When there is a new sampling point, X □ and X engineering. , and when the next new sampling point is between X and Xi, the counter is stopped and xL+!
When it moves between and xL+1, advance the counter by two,
When it moves between X□0 and Xl+-2, change the counter to
Proceed one step as usual.

The counter may be stopped at the time of enlargement, and the counter may be incremented by two at the time of reduction. That is, when enlarging, the position of the new sampling point is determined by the operation of incrementing the counter by one and the operation of stopping the counter. When reducing by 6, the position is determined by a combination of an operation that advances the counter by one and an operation that advances the counter by two. As long as reduction is considered within the range up to 50%, it is sufficient to advance the counter by one or two, but if the reduction rate is 50% or more, the counter may be advanced by three.

Information about where and how many steps the counter should be incremented.

It is calculated in advance by the CPU based on the magnification. For the new sampling point digit II x t, if the start position is O9, the old sampling pitch is 1, and the magnification is α (%), then X, = □xi (i = 0, 1...)... (9) α becomes.

Assuming that the new sampling point is between xl and Xi, the integer part of Xi in this case is i. That is,
As i increases, when the integer part of Xi increases by 1, the counter also advances by 1; when i increases and the integer part of Xi increases by 2, the counter also advances by 2, and the integer part of XL increases by 1.
If it does not advance at all, the counter should also not advance.

Further, the decimal part of Xi is the distance ΔX between xl and X.

This distance data will be used later in the interpolation calculation section.

The CPU calculates i=o to α-1 using equation 1 (9). This is because in all cases, the new sampling points are every α period. This calculation is performed after the magnification α (%) is specified before the start of the reading operation, and is written in a form that matches the hardware in a RAM or the like, and is sequentially read out during the scaling process.

Alternatively, as another method, a dedicated CPU or arithmetic means is provided, the above formula (9) is calculated in parallel with the scaling process, and the integer part of the calculation result Xi is used as an address,
It is also possible to use the decimal part as distance data.

An example of the configuration of the data synthesis section and line memory section is shown in FIG. This example is a configuration example when two-pixel correction is performed.

Here, when the RAM is in the input state, the address generation section advances the address counter in synchronization with DCLK in normal operation, but when the RAM is in the output state, it modifies the address by the method described below.

The first method is to change the frequency of the counter clock. If the frequency of DCLK is fo,
The frequency f when changing α (%) is f = −−f a
...(10) α becomes. This method is accurate and reliable because the deviation of fCx with respect to fo is the deviation of the sampling point itself.

When reading from the RAM, desired composite data can be obtained by moving the address counter at the above-mentioned f and sampling (latching) the output of the RAM again at DCLK.

With this method, the calculation result of equation (9) mentioned earlier is
Information about the integer part is no longer necessary, and only information about the decimal part, that is, information about the distance is required.

Another method is to focus on the integer part of the calculation result of equation (9), and use When the integer part increases by 2 → aL=0 (2) When expanding: When the integer part increases by 1 → al=
1 Define a sequence (ai) from i=0 to α-1 such that when the integer part is not increasing → ai=0, and store it in RAM
There is a method in which a clock having the frequency 2f0, which is twice the frequency of DCLK, is prepared as a clock.

During magnification processing, the device output consisting of RAM is j
It is assumed that =O to α-1 are read out repeatedly. During reduction, the clock of the address counter for the output of the line memory (RAMI or RAM2) is switched so that when a = 1: fo (DCLK) and when a = 0: 2f0.

When expanding, the address counter clock is a and DCL.
By ANDing K, the count is up when a = 1, and the count is not counted when a = 0.

A block diagram of a circuit for implementing this method is shown in FIG. 10. RAM 3 is a memory that stores ai and decimal part information of the result of equation (9) calculated by the CPU.

If this method is used, the data synthesis section shown in FIG. 8 can be omitted. In other words, the peripheral data can be obtained by inputting 6-bit data as is to the line memory in synchronization with DCLK, and after output, latching it in several stages using the aforementioned clock in the interpolation calculation section. It's for a reason.

Yet another method is shown in FIG. The address counter itself continues to count up based on DCLK. and,
Apart from the address counter, there is another one, this is UP/DO
A WN counter is provided, DOWN when enlarging and UP when reducing.
Make it so. The clock of this UP/DOWN counter is set to DC so that it counts only when ai=0.
Insert AND of LK and 1i.

By this, for example, when reducing, first ai=o
Set the UP/DOWN counter to 1 and add 1 to the value of the address counter to obtain the address of RAM1 or RAM2. Furthermore, at the next step a = 0, the UP/DOWN counter is set to 2, the address counter is added, and so on to determine the position of the new sampling point. In the case of expansion, conversely, a1=o and subtracts one by one, so UP/
The DOWN counter is decremented.

Next, a method of interpolation calculation will be explained.

The interpolation calculation involves calculating the above-mentioned equation (5), but the problem here is the positional accuracy of the new sampling point. In other words, the question is what accuracy should be considered for Δxi/p.

In the present invention, since a deviation of 173 dots between R, G, and B is corrected, an accuracy of about 0.1 dot is required. Therefore, it is sufficient to consider ΔXL/P with an accuracy of 1/8 or 1/16 dot. Furthermore, if emphasis is placed on correcting the positional deviation of 173 dots, it is effective to handle ΔX/p with an accuracy of 173 dots or 176 dots. That is, by shifting the new sampling points of the two colors on both sides of the new sampling point of the pixel of the central color among R, G, and B by ±173 dots,
Interpolated data can be obtained without calculation errors.

Here, as a representative example, a case will be explained in which processing is performed with an accuracy of 178 dots.

When handling with 178 dot precision, the decimal part bi of ΔX/p will be handled as 3-bit data. Third
In the figure, 8 sets of coefficients of equation (5) are prepared for 68 ways of quantizing Δx, /p in the 8 ways shown in FIG. Here, interpolation using 6-pixel reference with symmetry in mind will be explained.

In FIG. 16, for each of the above, pixel X□(i = -3
, -2, -1, 0, 1, 2).

FIG. 12 shows an example of the configuration of the interpolation calculation circuit. This circuit uses a latch to receive 6 pixel data output from the line memory shown in FIG.
The calculation of equation (5) is performed using a multiplier and an adder. On this occasion,
The coefficients must be changed as shown in FIG. 16 depending on the value of b.

In the example shown in FIG. 12, a ROM is used instead of a single multiplier to perform multiplication using a table reference formula and add two pixels. Furthermore, by inputting the above bi to the ROM address in synchronization with DCLK, the coefficients can be easily changed. The outputs from the ROM are added by two adders, and the interpolation operation is completed. here,
Although any two pixels may be selected for the first-stage addition using the ROM, it is possible to perform calculations with higher accuracy using a smaller number of bits by combining pixels with similar coefficient values as shown in FIG.

Here, 5O-S-□, S□-S-, , S2-S-.

It is good to combine.

The above is the calculation for the basic color component image. For the other two colors, it must be shifted by 173 dots. For example, this has an address generator for each color as shown in FIG. 10, and a RAM corresponding to each color.
3, when shifted by a predetermined amount, CP
All you have to do is calculate it with U and write it. However, since the amount of shift is always constant, it is possible to use only one address generator.

Next, an interpolation calculation method that is a further improvement on the above method will be described.

If the set of coefficients shown in FIG. 16 is selected, interpolated data can be obtained in units of 178 dots. Therefore, when the decimal part bi of the coordinates of the 8 standard colors, which can also be used to obtain interpolated data at a position shifted by 378 dots from the 173 dot valve, is 4 (more precisely, 4/8), for other colors, b, = If 4±3 is selected, interpolated data at a position ``+3/8 dot shift 1 - shift'' can be obtained.

3/8 = 0.375.1/3 = 0.3
33, and in practice, when the 0 position extreme degree is set to 174 dots, a sufficient positional deviation correction effect can be obtained even if 3/8 = 1/3.

Although a certain degree of effect can be obtained by shifting by ±174 dots, if emphasis is placed on positional deviation correction, 1/3 dot correction is more effective.

Returning again to the case of 178 dot accuracy. When bl is 2 or less or 5 or more, bi of other colors may become negative or exceed 7 due to a shift of ±3/8 dots. In the standard color, the interpolation position X exists between X-1 and xo, but due to the shift operation, in other colors,
This means that it may be between xl and xl. As a countermeasure for this, for colors that shift by -1/3 dots, use S-4, s-, s-, s-1, s, s□ as reference pixels.
For +SZ 7 pixels, +173 dots shifted color, s +3, s +i, s +, t S @
＃SL? If you prepare 7 pixels of S 2 F S 3, and the B type is other than O~7, in the former case, S
−2 to S i + in the latter case, the first
Interpolation calculations may be performed using the appropriate set of coefficients shown in FIG.

At this time, the coefficient for St in the former case and S-1 in the latter case is 0.

Fig. 17 shows S-4 to -173 dot shift.
In the case of +1/3 dot, which indicates the coefficient for S2, S
-1 to S, coefficients can be similarly determined.

FIG. 13 shows an example of the configuration of the arithmetic circuit in this case. Here, bi may be a signal common to the reference color, and the content of the ROM table may be the result of calculation using the coefficients shown in FIG. 17.

Note that in FIG. 17, Cnl1 indicates a coefficient when bl=n and pixel number=m.

Here are some other examples.

It has already been explained that by using the filter shown in FIG. 2(g) above, data shifted by -1/3 dots can be obtained. Here, by using coefficients obtained by superimposing the filter coefficients shown in FIG. 2(g) on the coefficients shown in FIG. 16, the shift operation and interpolation calculation are performed simultaneously. This combination results in a one-dimensional filter operation with sizes of 6×1 and 5×1, resulting in a 10×1 filter operation. Although 10 pixels may be referred to, in order to simplify the hardware, sufficient effects can be obtained by using approximately 6 to 8 pixels with large coefficients.

FIG. 18 shows coefficients when performing interpolation calculations with reference to seven pixels S-3 to S3. Coefficients can be found in the same manner for the case of +1/3 dot shift.

Here, we calculated new coefficients using two sets of filter coefficients, which is shown in Figure 15 as b = 0 to 71.
For ΔXa/P”Δx a / p + 1 / 3
This is the same as finding the coefficient from equation (5).

Both of the above two methods can use the data of ai + b in common for each color, so in particular ai, t)
The method of calculating i by the CPU can reduce the load on the CPU.

The combination of magnification processing and positional deviation correction processing has been described above in detail, but as described above, it is possible to further incorporate MTF correction into one. These processes can be combined into one by multiplying the coefficients for calculation, as shown in FIGS. 16 to 18. FIG. 19 shows the coefficients of the filter obtained by superimposing the MTF correction filter in the main scanning direction shown in FIG. 2(e) on the variable magnification and ten-position shift correction filter shown in FIG. be. Here, the reference pixels are 7 pixels.

Note that it is also possible to implement scaling and MTF without superimposing the three filters, and to use the above-mentioned improved method for positional deviation.

Effects As described in detail above, according to the present invention, in a color image processing device having an image reading means using a one-dimensional color image sensor in which color separation filters are regularly arranged, different colors are used for the reference color. Image shift processing includes means for obtaining image data when shifted by a predetermined shift amount by interpolating with reference to data of a plurality of pixels, and means for performing scaling processing at least in the main scanning direction. Since the present invention is constructed so that the image processing and magnification processing are performed simultaneously, it is possible to realize a color image processing method capable of correcting pixel positions without deteriorating the sharpness of one image, which is a remarkable effect.

[Brief explanation of the drawing]

1(a) to 1(d) are block diagrams of a digital color copying system showing an embodiment of the present invention, FIG. 2 is a diagram showing the structure of a digital filter, FIG. 3 is an explanatory diagram of sampling positions, and FIG. The figure is a graph showing the Sinc function, Figure 5 is a graph showing the MTF characteristics of the filter used in the example, and Figure 6 is a graph showing the MTF characteristics of the filter used in the example.
7(a) is a block diagram showing an example of the circuit configuration of a positional deviation correction filter.
Figure (b) is a block diagram showing an example of the configuration of the MTF correction circuit in the sub-scanning direction, Figure 8 is a block diagram showing an example of the configuration of the scaling processing circuit, and Figures 9 to 11 are the data synthesis section and line memory. FIGS. 12 and 13 are block diagrams showing examples of the configuration of a correction calculation circuit; FIGS. FIG. 14 is a diagram explaining the digital filter along with its calculation process, FIG. 15 is a diagram showing the process of positional deviation correction,
Figures 16 to 19 are diagrams showing the coefficients of the synthesized digital filter, Figure 20 is a diagram showing the line sensor, and Figure 2 is a diagram showing the coefficients of the synthesized digital filter.
FIG. 1 is a diagram showing the state of sampling by a line sensor, and FIGS. 22 to 25 are diagrams for explaining problems in the prior art. input system, 2: shading correction circuit, 3: positional deviation correction circuit, 4: MTF correction circuit, 5: variable magnification circuit, 6: γ
Correction circuit, 7: Color correction/UCR circuit, 8: Gradation processing circuit, 9: Output system, lO: Positional deviation correction/magnification circuit, 11:
Positional deviation correction/MTF correction/variable magnification circuit. Patent applicant Rico Co., Ltd. - Figure 2 (a) (b) (
c) Figure 2 (h) Figure 3 Nutrient diagram:! l! Figure 7 (&) Latch x 5 Figure 7 (b) Line memory x 2 Figure 10 Reduced figure 11 Figure Kyo Figure 15 Figure 20 Figure 21 Figure GE50ER Figure 22 Figure 23 (Area ratio equivalent %) Figure 24 (Area ratio equivalent '5) Boiled H

Claims (4)

[Claims] (1) In a color image processing device having an image reading means using a one-dimensional color image sensor in which color separation filters are regularly arranged, image data obtained when another color is shifted by a predetermined shift amount with respect to the reference color is , comprising means for obtaining by interpolating with reference to data of a plurality of pixels, and means for performing scaling processing at least in the main scanning direction,
A color image processing method, characterized in that it is configured to perform image shift processing and scaling processing at the same time. (2) The color image processing method according to claim 1, wherein the scaling processing means uses an interpolation formula based on a Sinc function. (3) The color image processing method according to claim 1 or 2, wherein the interpolation is performed using a Sinc function. (4) The color image processing method according to claim 1, 2, or 3, wherein the number of reference pixels when performing the interpolation is 4 or more pixels.