JP2658087B2 - Color image processing method - Google Patents

Description

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 or the like.

2. Description of the Related Art In a document in which pictures and characters are mixed, characters are often black even in a full-color document. To reproduce black with a full-color copying machine, three color materials of Yellow, Magenta, and Cyan (hereinafter, referred to as “Y, M, C”) are superimposed.
However, even if the three colors Y, M, and C are superimposed, some color components remain if the balance between Y, M, and C is not completely achieved. Further, if the alignment of the Y, M, and C plates is incomplete and misaligned, the quality of an image such as a character that requires high resolution is significantly degraded. As a countermeasure against this problem, digital color copiers use black (B
The application of UCR (under color removal) to replace with K) is considered.

On the other hand, as an example of a color image input method, for example, a method of obtaining a color signal by forming an image on three CCDs using a dichroic prism disclosed in JP-A-60-187180 is known. . In this method, it is difficult to completely eliminate the chromatic aberration of the image forming optical system, and one point of the original is accurately determined for each point.
An image is not formed at a predetermined position of the CCD, and therefore, the color balance of the read pixel signal is out of order. There are also disadvantages such as difficulty in adjusting the position of the CCD, a large number of parts, and high cost.

In another example, Japanese Patent Application Laid-Open No. 61-61561 discloses that a red, green, blue (hereinafter, referred to as "R, G, B") filter is regularly arranged on the elements of the photoelectric conversion array to adjust the position. In addition, there is one that solves the problem of the number of parts. This method has another problem that, since the R, G, and B photoreceptors are displaced from each other, data at the same point of the original image cannot be read, and the color correction processing and the UCR processing cannot be performed accurately. It had the following.

To address this problem, the present applicant has previously filed Japanese Patent Application No. 61-19 / 1986.
There is a method proposed by No. 4712 “Color image processing method”. This method uses the above-described image reading apparatus using a one-dimensional color image sensor in which color separation filters are regularly arranged, in order to read the position of an image with high accuracy, to shift the dot position between R, G, and B. Is linearly interpolated using the data of two adjacent pixels to obtain image data at a predetermined position. This method can correct the pixel position, but performing linear interpolation means taking a weighted average, which has a low-pass characteristic as a digital filter. However, there is a problem that the sharpness of a line drawing or the like is reduced.

Hereinafter, this will be described in detail. FIG. 20 shows G,
1 shows a part of a one-dimensional color image sensor (line sensor) in which B and R color separation filters are regularly arranged. In the figure, each pixel indicated by n−4 to n + 3 is subdivided into G,
It is composed of three small pixels having 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 1/3 pixel between GB, BR and 2/3 pixel when viewed between GR. Consider the effect of this displacement on the UCR process. Specifically, the sampling density is about 16 dots / mm,
As an image, especially for black characters where UCR processing is a problem,
Since the thickness of the line is about 100 μm, the state of sampling in this case is as shown in FIG.

Here, in order to simplify the description, it is assumed that the output value of the image is a linear reflectance, and the color material, paper, and the like are ideal. 21st
In the figure, the black line extends over pixels n and n + 1. In an ideal sensor having no displacement between the colors, the output values (reflectance%) of the pixels n and n + 1 are 50% and 17% for G, B and R, respectively. When this is reproduced using M, Y, and C color materials that are complementary colors of G, B, and R, the area ratio is 50% and 83%, respectively.
Color materials should be used.

Further, if the UCR 100% processing is performed, the minimum values of M, Y, and C can be replaced with BK, so that only BK needs to be used in 50% and 83%. However, in reality, since there is a displacement between R, G, and B, the output values of each color are as shown in FIG. Therefore, the necessary M, Y, and C are as shown in FIG. 23.
The amounts of the necessary color materials of Y, C and BK are as shown in FIG. 24, and it can be seen that it is impossible to reproduce with one color of BK.

In the Japanese Patent Application No. 61-194712, "Color Image Processing Method", linear interpolation is performed using data between two adjacent pixels in order to correct a positional shift between colors. Considering the sensor position of B color as the center, G and R colors are respectively -1 /
There is a displacement of 3 pixels and 1/3 pixel. When this is corrected by linear interpolation, the following equation is obtained.

Here, G _n , R _n−1, etc. represent output values of the n-th pixel of G color and the n−1-th pixel of R color, and G _n ′, B _n ′, R _n ′ represent correction values .

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

When the filter characteristic is represented by f (x), its DFT is u: spatial frequency N: 1 / Δx Δx: sampling pitch m: filter size MTF (u) = | F (u) | (3)

FIG. 25 shows the MTF characteristic of the digital filter when F (x) = [1 / 3,2 / 3,0] m = 3 Δx = 62.5 μm (16 dots / mm) (f
(X) = [0,2 / 3,1 / 3] also has the same characteristic). As is clear from the figure, this filter has poor transmission of high-frequency components and has a low-pass characteristic.

Therefore, by performing the displacement correction using this filter, the high frequency components of the G and R color images are degraded. That is, the G and R images are blurred, and the sharpness of characters, line drawings, and the like is reduced.

The present invention has been made in view of the above circumstances, and aims to solve the above-described problems in the conventional technology and correct the pixel position without deteriorating the sharpness of an image. An object of the present invention is to provide a color image processing method which can be obtained.

The object of the present invention is to provide a color image processing apparatus having image reading means using a one-dimensional color image pickup device in which color separation filters are regularly arranged, wherein another color is shifted by a predetermined shift amount with respect to a reference color. Image data,
This is achieved by a color image processing method characterized in that it is obtained by interpolating with reference to data of a plurality of pixels.

Hereinafter, after describing the principle of the present invention, the configuration will be described in more detail based on embodiments.

According to the sampling theorem, the one-dimensional signal g (t) band-limited to the frequency W or less is sampled at an interval of T = 1 / (2W), that is, at a sampling frequency of f _s = 2W. Thus, the original signal can be completely restored.

According to the color image processing method according to the present invention, interpolation data is obtained by applying the principle of the above equation (4) to the sampled image, and misregistration is corrected.

Here, a method of calculating the interpolation data will be described. FIG. 3 shows the relationship between the positions of the sampling pixels, the positions for obtaining the interpolation data, and other parameters. Using the parameters shown in FIG. 3, the interpolation data O (x) can be obtained by the following equation.

Where h (x) is an interpolation function, Also, Is a normalization coefficient. i is from −∞ to ∞, but as can be seen from the graph of h (x) shown in FIG. 4, | x |
Is larger, h (x) → 0. Therefore, in practical use, i may be in the range of about -2 to 2.

By the way, the interpolation data may be obtained for the remaining two colors based on any one of R, G, and B. Here, a case will be described in which positional deviation correction is performed on G and R with reference to B. However, the positional relationship between R, G, and B is as shown in FIG. Ie, R data will be used interpolated data at the shifted 1/3 pixel left of the target pixel position x ₀ position.

Here, assuming that x ₀ is the origin (x ₀ = 0), the position for obtaining the interpolation data is It is. By using the above equation (5), the interpolation data is as follows.

FIG. 14 shows the value of each coefficient when the data of five pixels of i = −2, −1, 0, 1, 2 is used. One-dimensional filters using the coefficients shown in FIG. 14 [10.1676, 0.4191, 0.8383, −0.20
95, 0.1197], an image shifted by 1/3 pixel can be obtained. That is, the image can be converted into an image having the same phase as the B color.

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

One thing to note when approximating coefficients to rational numbers is the ratio between the coefficients. If the coefficient for pixel i is C _i , Although some of the calculations in the middle are omitted, In a small range of | i |, | j |, that is, a large range of C _i , C _j , the coefficient ratio becomes a simple integer ratio. Thereby, the accuracy of the position (coordinates) for obtaining the interpolation data can be ensured.

In the above description, the interpolation of the R color data has been described. The filter coefficient can be calculated using equation (6). In addition, when the R color is used as a reference among G, B, and R, for B, For G, Thus, the same calculation can be performed using Expression (6). According to the method described above, a correction filter with little MTF deterioration can be obtained.

Hereinafter, embodiments of the present invention will be described in detail with reference to the drawings.

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

Hereinafter, the position shift correction circuit 3, the MTF correction circuit 4, the γ correction circuit
6, through the color correction / UCR circuit 7 and the gradation processing circuit 8, binarized (or multi-valued about 3 to 8 values) image data is sent to the output system 9, and a copied image is obtained.

Here, the MTF correction circuit 4 corrects blurring of an image due to an input system. 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 from the MTF characteristics of the actual input system, but is not limited to 3 × 3, and a filter having a size of 5 × 3, 5 × 5, etc. may be used. Further, as shown in FIGS. 7E and 7F, the correction can be made separately in the main scanning direction and the sub-scanning direction.

By the way, the position shift correction circuit according to the present invention is a one-dimensional filter, and it is possible to configure one filter in combination with the next stage MTF correction filter. FIG. 2 (h) shows a combination of the MTF correction filter (a) and the positional deviation correction filter (g). (I) is a combination of an MTF correction filter (e) in the main scanning direction and a positional deviation correction filter (g).

In FIG. 2 (h), the size of 7 × 3 is obtained by combining filters of 3 × 3 and 5 × 1, and in FIG. 2 (i), the size of 7 × 3 is obtained by combining 3 × 1 and 5 × 1 filters. The filter has a size of × 1.

Here, the coefficient outside the filter may be omitted in consideration of being smaller than the coefficient near the center. FIG. 2 (j) shows a 5 × 1 filter obtained by ignoring the two outer elements of FIG. 2 (i). The same simplification can be made for (h). By combining filters as described above, hardware can be simplified and calculation accuracy can be improved.

FIG. 1 (b) shows a system configuration in which the position shift correction circuit and the MTF correction circuit are integrated into one.

FIGS. 1 (c) and 1 (d) show a system in which a scaling circuit 5 is added to the system shown in FIG. 1 (a). In a digital copying system, a wide range of magnification can be obtained relatively easily, not only when an equal-magnification optical system is used as an input system but also when a reduction optical system is used. Is used.

The electrical scaling method converts a pixel data string sampled at the same magnification into a data string having a pixel density corresponding to the scaling factor. For example, when data of 100 pixels is sampled at the same magnification, when 120% enlargement is performed, 120 pixel data is obtained based on 100 pixel data, and when reduced by 70%, 70 pixel data is obtained. Perform the operation that requests.

As an example of the above-described electric zooming device, the applicant of the present invention has previously disclosed Japanese Patent Application Nos. 61-100503, 61-100505, and 61-1005.
Nos. 06, 61-101721, 61-104014, and 62-32
There is a device proposed in No. 62.

Among the electric scaling methods, those using a Sinc function as an interpolation function can accurately obtain scaling data.

FIG. 6 shows the sampling position (x _i ) of the same-size data,
The relationship between the sampling position (y _i ) at 133% enlargement and the sampling position (z _i ) at 80% reduction is shown. In this scaling method, image data at a position indicated by y _i and z _i is obtained by using the data at a position indicated by x _i according to the above-described equation (5).

By changing the density of the sampling points of the scaled image indicated by y _i and z _i , it is possible to change the magnification at an arbitrary scale.

In this embodiment, by shifting the new sampling position after the above-described scaling by a predetermined amount for each color, the positional deviation between the R, G, and B pixels is corrected simultaneously with the scaling processing. . That is, considering the relationship between the sampling positions of the three colors in the case of FIG.
This is obtained by interpolating pixel data at a position shifted by 1/3 pixel to the right for 1/3 pixel and G color.

FIG. 1 (d) shows a digital copying system having a scaling unit. In this embodiment, as shown in FIG. It is also possible to combine both processing units into one.

As described above, the position shift correction processing can be combined with the MTF correction processing or the scaling processing. The combination with the MTF correction processing is to superimpose the function for MTF correction on the interpolation function shown by the equation (5). It is to shift the position of the sampling to be obtained.

Therefore, if these two combinations are performed at the same time, it is possible to combine the three processing units of the positional deviation correction, the MTF correction, and the magnification change into one. Fig. 1 (f) shows the position shift correction and MTF
1 shows a configuration example of a copying system having a processing unit in which three processing units of correction and scaling are combined. In this case, the MTF correction circuit in the subsequent stage performs MTF correction in the sub-scanning direction. Although it is possible to combine them into one including the MTF correction in the sub-scanning direction, it is advantageous to separate them to simplify the correction calculation unit.

 Hereinafter, a specific configuration example will be described.

FIG. 7 (a) is an example of a circuit configuration of the displacement correction filter shown in FIG. 2 (g). By receiving the image data output from the input system by the five-stage latches, data of five pixels can be simultaneously referred to in the main scanning direction. In this example, the multiplication of the image data and the coefficient of the displacement correction filter shown in FIG. 2 (g) is performed in a table reference formula using a ROM instead of the multiplier.

Further, S ₀ and S ₋₁ and S ₁ and S ₋₂ are paired, and the first-stage addition is performed simultaneously. That is, the memory address is the address in the image data of the S ₀ and S _-1 or S ₁ and S _-2, sum of products of the coefficients corresponding to the respective image data are stored. The intermediate result of the operation output from ROM is
The addition is performed by the next adder, and the operation ends.

A filter combining position shift correction and MTF correction in the main scanning direction (see FIGS. 2 (i) and 2 (j)) can be realized by extending the circuit of FIG. 7 (a). That is, the required number of latches and the necessary number of multipliers and adders at the subsequent stage may be added. As shown in FIG. 7A, the difference between the coefficients is that when the ROM table method is used, only the data stored in the ROM needs to be changed.

FIG. 7B is a configuration example of the MTF correction circuit in the sub-scanning direction. By sequentially updating and storing the image data in the line memories for two lines, it is possible to simultaneously refer to the data of the corresponding three pixels of three consecutive lines.

Here, the MTF correction filter of FIG. 2 (f) will be described. Data of two peripheral pixels corresponding to a coefficient of −1 are added by an adder. The ROM at the next stage is used to calculate the data of the center pixel and the two surrounding pixels. The calculation of subtracting the output value from the adder from the value three times (the value of the center pixel) is performed by using a table reference formula. Is what you do.

As described above, when performing the MTF correction separately in the main scanning direction and the sub-scanning direction, either may be performed first. That is, whichever of the circuits shown in FIGS. 7A and 7B may be provided first.

 Next, the scaling process will be described.

The scaling process is performed, for example, by a circuit having a configuration as shown in FIG. The circuit shown in FIG. 8 has a function of determining the position of a new sampling point after zooming, and a function of calculating new data from the distance between the new sampling point and the old sampling point and the old data. .

First, in the future, when a new sampling point is determined and an operation is performed in the future, the data synthesizing unit extracts the peripheral data at one time.
This is to be summarized for each pixel. For example, in FIG. 3, if the new sampling point x is between x ₀ and _{x -1, S -3, S -2} , S -1, S 0, S 1, S 2 at a time by the arithmetic unit It is to take out.

A specific method can be implemented by latching data (DATA1) sequentially input in synchronization with the data clock with the data clock (DCLK). For 6 pixels,
This can be realized by a five-stage latch.

Next, a line memory unit is a memory for storing a group of 4 to 8 pixels as many as the number of pixels in one line. The memory has a two-stage configuration of input and output. The output and input / output are reversed when one line ends.

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

A new sampling point is, at one time, between x _i and x _{i + 1} , and the next new sampling point is once again x _i and x _{i + 1}
When it is between, the counter is stopped, when it moves between x _{i + 2} and x _{i + 3} , the counter is advanced by two, when it moves between x _{i + 1} and x _{i + 2} , , Increment the counter by one as usual.

The counter is stopped at the time of enlargement, and the counter is set to 2
The next step can be considered at the time of contraction. That is, at the time of enlargement, the position of the new sampling point is determined by an operation of incrementing the counter by one and an operation of stopping the counter. At the time of reduction, the position is determined by a combination of an operation of advancing the counter by one and an operation of advancing the counter by two. As long as the reduction is considered within the range up to 50%, the counter may be advanced by one or two, but in the case of a reduction ratio of 50% or more, the counter may be advanced by three.

Information on where to advance the counter and how many times the counter is advanced is calculated in advance by the CPU according to the magnification. The position x _i of the new sampling point is 0 for the start position and 1 for the old sampling pitch.
And the magnification is α (%), Becomes

If the new sampling point is between x _i and x _{i + 1} ,
In this case, the integer part of x _i is i. That is, when the integer part of x _i increases by one with the increase of i, the counter is also advanced by one, and when the integer part of x _i increases by two with the increase of _i ,
The counter is also advanced by two, and if the integer part of x _i does not advance by one, the counter may not be advanced.

Further, the fractional part of x _i will distance Δx between x _i and x. This distance data will be used in the interpolation calculator later. CP
In U, i = 0 to α−1 is calculated by equation (9). This is because, in all cases, a new sampling point has a period of every α points. This calculation is performed before the start of the read operation.
This is performed after the magnification α (%) is specified, and written to RAM, etc. in a form that matches the hardware.
Read sequentially.

Alternatively, as another method, a dedicated CPU, or the arithmetic means is provided, in parallel with the scaling process to calculate the equation (9), the calculation result as the integer part of x _i, and the address, the fractional part It can be used as distance data.

FIG. 9 shows a configuration example of the data synthesizing unit and the line memory unit. This example is a configuration example when performing two-pixel correction. Here, when the RAM is in the input state, the address counter advances in synchronization with DCLK in the normal operation, but when the RAM is in the output state, the address generator corrects the address by the method described below.

The first method is to change the frequency of the clock of the counter. The frequency of DCLK, and the f _0, α
(%) The frequency f _α during zooming is Becomes In this method, since the deviation of the f _alpha for f ₀ is shifted itself sampling points, it is accurate and reliable.

When reading RAM, move the address counter in the above f _alpha, by sampling (latching) again DCLK output of RAM, it is possible to obtain a desired synthetic data.

According to this method, information on the integer part is unnecessary in the calculation result of Equation (9) described above, and only information on the decimal part, that is, information on the distance is required.

Another method focuses on the integer part in the calculation result of equation (9), and uses x _i and x _{i + 1} to: (1) Reduction: when the integer part increases by one → a _i = 1 When the integer part increases by two → a _i = 0 (2) When expanding: When the integer part increases by one → a _i = 1 When the integer part does not increase → a _i = There is a method in which a sequence {a _i } of 0 is defined from i = 0 to α−1, written in the RAM, and the DCLK and a clock having a frequency 2f ₀ which is twice this frequency are prepared as a clock.

During the scaling process, a _i is read from the RAM. Read i =
It is assumed that 0 to α-1 are repeatedly read. When shrinking,
The clock of the address counter for the output of the line memory (RAM1 or RAM2) is switched so that when a _i = 1: f ₀ (DCLK) when a _i = 0: 2f ₀ .

At the time of enlargement, the address counter clock is _ai and DCLK
By performing an AND operation, count up when a _i = 1 and do not count when a _i = 0.

FIG. 10 is a block diagram of a circuit for implementing this method. The RAM 3 is a memory that stores _ai and the information of the decimal part of the result of the equation (9) calculated by the CPU. When this method is used, the data synthesizing unit shown in FIG. 8 can be omitted. That is, the line memory stores 6-bit data as it is,
This is because peripheral data can be obtained by inputting the data in synchronization with DCLK, outputting the data, and then latching several stages with the above-described clock in the interpolation operation unit.

Yet another scheme is shown in FIG. The address counter itself continues counting up with DCLK. In addition to the address counter, another one is provided with an UP / DOWN counter, which is set to DOWN for enlargement and UP for reduction. As for the clock of the UP / DOWN counter, AND of DCLK and _ai is inserted so as to count only when _ai = 0.

Thus, for example, at the time of reduction, first, the first a _i = 0
Then, the UP / DOWN counter is set to 1, and the value of the address counter is incremented by 1 to obtain the address of RAM1 or RAM2. Furthermore,
At the next a _i = 0, the UP / DOWN counter is set to 2, and the address counter is added to determine the position of the new sampling point. In the case of enlargement, the UP / DOWN counter is decremented in order to subtract one by one at _ai = 0.

 Next, a method of the interpolation calculation will be described.

The interpolation calculation is to perform the calculation of the above equation (5), but what matters here is the positional accuracy of the new sampling point. That is, how much accuracy Δx _i / p is considered.

In the present invention, since the deviation of 1/3 dot between R, G and B is corrected, an accuracy of about 0.1 dot is required. Therefore, Δx _i / p may be considered with an accuracy of 1/8 or 1/16 dot. Further, if emphasis is placed on the correction of the positional shift of 1/3 dot, it is effective to handle Δx _i / p with an accuracy of 1/3 dot or 1/6 dot. That is, the new sampling point of the pixel of the color at the center position of R, G, B
By shifting the new sampling point of the color by ± 1/3 dot, the interpolation data can be obtained without a calculation error.

Here, a case of handling with 1/8 dot precision will be described as a representative.

When handling with 1/8 dot precision, the decimal part b _i of Δx _i / p
Is treated as 3-bit data. In FIG. 3, Δx ₀ / p is quantized as shown in FIG. Eight of b _i
, Eight sets of coefficients of the equation (5) are prepared. here,
Interpolation with reference to six pixels in consideration of symmetry will be described.

For each b _i in FIG. 16, the pixel _{x i (i = -3, -2} , -
1,0,1,2).

FIG. 12 shows a configuration example of the interpolation operation circuit. In this circuit, the data of six pixels referred to at the time of interpolation, which is output from the line memory shown in FIG. 9 described above, is received by the latch, and the operation of equation (5) is performed by the multiplier and the adder. In this case, it must be changed as the coefficients of FIG. 16 by the value of b _i.

In the example shown in FIG. 12, a table reference expression is multiplied by using a ROM instead of the multiplier, and two pixels are added. Furthermore, the above b is synchronized with the DCLK at the address of the ROM.
By inputting _i , the coefficient can be easily changed. The outputs from the ROM are added by the two adders, and the interpolation operation ends. Here, in the first-stage addition using the ROM, any two pixels may be selected, but it is more accurate to combine two or more pixels having the same coefficient values in FIG. 16 with a smaller number of bits. Can be. Here, S ₀ −S
_It is better to combine _-1 , S ₁ -S _-2 and S ₂ -S _-3 .

The above is the calculation for the basic color component image. The other two colors must be shifted by 1/3 dot. This is because, for example, the address generator shown in FIG. 10 is provided for each color, and the RAM 3 corresponding to each color has
It is sufficient that the CPU calculates and writes a _i and b _i after shifting by a predetermined amount. However, since the shift amount is always constant, it is possible to use only one address generator.

Next, a description will be given of an interpolation calculation method which is a further improvement of the above method.

If a set of coefficients shown in FIG. 16 is selected, interpolation data can be obtained in 1/8 dot units. Therefore, 1/3 dot ≒ 3/8
Interpolation data at the dot shifted position can also be obtained. Fractional portion b _i of the reference color coordinates 4 (to be precise, 4/8)
At this time, if b _i = 4 ± 3 is selected for other colors, interpolation data at a position shifted by 3/8 dot can be obtained.

3/8 = 0.375, 1/3 = 0.333, and 3/8 = 1/3 in practical use
Thus, a sufficient positional deviation correction effect can be obtained. When the position accuracy is 1/4 dot, a certain effect can be obtained by shifting by ± 1/4 dot. However, when emphasizing the positional deviation correction, 1/3 dot correction is more effective.

Again, return to the case of 1/8 dot precision. When b _i is 2 or less or 5 or more, the other color b
There are cases where _i is negative or exceeds 7. In the reference color, the interpolation position x exists between x _-1 and x _0, but the shift operation, in other colors, is when x comes between x _-2 and x _-1, x ₀ and x ₁ There will be. As a countermeasure, S _-4 , S _-3 , S _-2 , S
_-1 , S ₀ , S ₁ , S _{2 For} 7 pixels, +1/3 dot shifted color, S _-3 , S _-2 , S _-1 , S ₀ , S ₁ , S ₂ , S ₃ of previously prepared seven pixels, if b _i becomes non-0 to 7, S _-4 to S ₁ in the former, the latter suitable coefficients of Figure 16 to 6 pixels of S _-2 to S ₃ What is necessary is just to perform an interpolation operation using a set. At this time, the coefficient for S _{2 in} the former and S _{-3 in} the latter is 0.

FIG. 17 shows coefficients for S _{-4 to} S ₂ when shifting by −−1 dot. In the case of +1/3 dots, the coefficients can be similarly determined for S _{-3 to} S ₃ . FIG. 13 shows a configuration example of the arithmetic circuit in this case. Here, b _i uses a common signal with a reference color, it is sufficient to the operation result using a coefficient indicating the contents of the ROM table in FIG. 17.

In FIG. 17, C _nm is b _i = n, pixel number = m
The coefficient at the time of is shown.

 Here are other examples.

As described above, it is possible to obtain data shifted by / 3 dot by using the filter shown in FIG. 2 (g). Here, the shift operation and the interpolation operation are performed simultaneously by using a coefficient obtained by superimposing the filter coefficient shown in FIG. 2 (g) on the coefficient shown in FIG. This combination is an operation of a one-dimensional filter of a size of 6 × 1 and 5 × 1, and eventually becomes a filter operation of 10 × 1. Although 10 pixels may be referred to, a sufficient effect can be obtained by using about 6 to 8 pixels having a large coefficient for hardware simplification.

In FIG. 18 shows a coefficient when performing interpolation calculation with reference to the 7 pixels of S _-3 to S _3. The coefficient can be obtained in the same manner in the case of +1/3 dot shift. Here, a new coefficient was calculated using two sets of filter coefficients. In FIG. 15, Δx _{0 is obtained} for b _i = 0 to 7.
This is the same as calculating the coefficient from equation (5), assuming that / p = Δx ₀ / p + 1/3.

Both of the above two methods can use the data of a _i and b _i in common for each color. Therefore, in the method of calculating a _i and b _i by the CPU, the load on the CPU can be reduced. .

As described above, the combination of the scaling processing and the positional deviation correction processing has been described in detail. As described above, it is possible to further incorporate the MTF correction into one. As shown in FIGS. 16 to 18, these processes can be integrated into one by multiplying by a coefficient for calculation. 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 magnification / position shift correction filter shown in FIG. is there. Here, the number of reference pixels is seven.

It is also possible to combine the magnification and the MTF without superimposing the three filters, and to use the above-described improved method for the positional deviation.

Effects As described in detail above, according to the present invention, in a color image processing apparatus having an image reading unit using a one-dimensional color imaging device in which color separation filters are regularly arranged,
Since the image data obtained when the other color is shifted by a predetermined shift amount with respect to the reference color is obtained by interpolating with reference to the data of a plurality of pixels, the pixel data can be obtained without deteriorating the sharpness of the image. This has a remarkable effect that a color image processing method capable of correcting the position can be realized.

[Brief description of the drawings]

1 (a) to 1 (f) are diagrams showing the configuration of a digital color copying system showing an embodiment of the present invention, FIG. 2 is a diagram showing the configuration of a digital filter, FIG. FIG. 5 is a graph showing the Sinc function, FIG. 5 is a graph showing the MTF characteristic of the filter used in the embodiment, FIG. 6 is a diagram for explaining the scaling process, and FIG. FIG. 7B is a block diagram showing a configuration example of an MTF correction circuit in the sub-scanning direction, FIG. 8 is a block diagram showing a configuration example of a scaling processing circuit, and FIGS. FIG. 11 is a block diagram showing a configuration example of a data synthesizing section and a line memory section, FIGS. 12 and 13 are block diagrams showing a configuration example of a correction operation circuit, and FIG. FIG. 15 is a diagram showing a process of position shift correction, FIG. 19 to 19 show the coefficients of the synthesized digital filter, FIG. 20 shows the line sensor, FIG. 21 shows the state of sampling by the line sensor, and FIGS. 22 to 25 show the prior art. It is a figure for explaining the problem of. 1: Input system, 2: Shading correction circuit, 3: Position shift correction circuit, 4: MTF correction circuit, 5: Magnification circuit, 6: γ correction circuit, 7:
Color correction / UCR circuit, 8: gradation processing circuit, 9: output system, 10: position shift correction / MTF correction circuit, 11: position shift correction / magnification circuit,
12: Position shift correction / MTF correction / magnification circuit.

Claims (5)

(57) [Claims] In a color image processing apparatus having image reading means using a one-dimensional color image sensor in which color separation filters are regularly arranged, an image obtained by shifting another color by a predetermined shift amount with respect to a reference color is provided. A color image processing method characterized in that data is obtained by interpolating with reference to data of a plurality of pixels. 2. The color image processing method according to claim 1, wherein said predetermined shift amount is ± (1/3) dots. 3. The color image processing method according to claim 1, wherein said interpolation is performed using a Sinc function. 4. The color image processing method according to claim 1, wherein the interpolation is performed by combining two pixels at a time with the largest weight. 5. The method according to claim 1, wherein the number of reference pixels at the time of performing the interpolation is four or more.
5. The color image processing method according to claim 2, 3, or 4.