JPS6038977A - Picture signal processing device - Google Patents

Abstract

PURPOSE:To display a color picture in an excellent way even with pattern noise such as a mesh point by applying spatial average to a picture signal obtained from a reader to obtain regulation, and applying binary-coding by means of a fixed threshold value when the regulation is large and applying binary-coding by means of a dither method when the regulation is smaller than a prescribed value. CONSTITUTION:A small memory capacity is enough for a color conversion table by deciding a coefficient aij so that an address subject to coordinate transformation is distributed almost uniformly on a coordinate space to each color, in a color copying machine where a picture is read by plural CCDs and conducting color output by means of a thermal transfer system. Further, the 2nd and 3rd coordinate transforming circuits 49b, 49c are constituted almost the same as the 1st coordinate transformation circuit 49a. The signal subject to coordinate transformation is applied with a averaging processing and a color conversion table is indexed by the averaged signal. Only when the local regulation of a luminance signal is larger than a prescribed value in this table, binary-coding is applied with the fixed threshold value and the dither method is applied in other cases. That is, a beautiful picture display is conducted by using the binary representation in the region where a picture signal fluctuates remarkably and using the dither expression in other regions.

Description

[Detailed description of the invention]

[Technical field of the invention] This invention converts a color image into an electrical signal using an image sensor,
The present invention relates to a device that outputs an image to a color display device using this electrical signal. In particular, the present invention relates to an image signal processing apparatus that can obtain a good image even when information that becomes a noise component such as halftone dots is superimposed on an input image. [Technical background of the invention and its problems] Conventionally, when an image of a printed matter containing characters and color images is photoelectrically converted using an image sensor and output to a color printer, etc., if inappropriate scanning is used to detect the image, the color image portion Moiré occurs between the halftone dots, making it difficult to process as a color image. In particular, when using a space-divided color sensor, it is almost impossible to reproduce the original color tone of halftone colors with halftone dots. That is, when emphasizing color difference signals to express an image with high saturation comparable to that of the input image, this halftone noise is also emphasized, resulting in a noisy image. In order to remove such noise, it is possible to obtain a blurred image using an optical system, but if this is done, even the letters become blurred, making it impossible to express the letters. It is also possible to average such a blurred image using electrical processing, but if such processing is simply applied to the sensor output, there are inconveniences such as the inability to display small characters. Ta. [Purpose of the Invention] This invention was made to eliminate the above-mentioned drawbacks, and it is possible to display a color image well even with pattern noise such as halftone dots, and it is possible to display a color image well even with pattern noise such as halftone dots. An object of the present invention is to provide a color human-powered display device capable of displaying images with high resolution even when the image is displayed. [Summary of the Invention] The present invention performs spatial averaging on an image signal obtained from a reading device, and calculates the fluctuation rate of this signal. Then, when the fluctuation rate is large, binarization is performed using a fixed threshold value, and when the fluctuation rate is smaller than a predetermined value, the dither method is used to perform binarization.
It is something that does value. [Effects of the Invention] This invention captures the image formed by the halftone dots, and when outputting and displaying the image, averages the image signal obtained from the glare filtering device, so the influence of halftone noise can be eliminated.
By switching the binary dither, a sharp image can be formed. [Embodiments of the Invention] Before describing the embodiments of the present invention in detail, an outline will be described. First, a color image is converted into an electric signal by an image sensor, and this electric signal is separated and converted into a luminance signal and a color difference signal or a signal equivalent thereto by a matrix circuit including a product-sum calculation circuit. The coefficients of the matrix are determined so that the signal is distributed uniformly over a wide range. Next, averaging processing is performed so that these signals have an SN above a certain level even if halftone noise is superimposed. In this case, it is generally necessary to average more color difference signals. Also, the local fluctuation rate is calculated for the luminance signal. Furthermore, the difference between the averaged luminance signal and the original luminance signal is taken, and the difference between the luminance signal and the original luminance signal is calculated. If the local variation rate determined in advance is larger than a predetermined value, binarization is performed using a fixed threshold, and in other cases is dithered and displayed. By doing so, it is possible to display even small characters with high resolution, and even for images containing pattern noise such as halftone dots, it is possible to display a good image with little influence of noise. Next, embodiments of the invention will be described with reference to the drawings. An embodiment with a plurality of CCDs (Charge Couple
The present invention relates to a color copying machine that reads images using an edDev system and performs color output using a thermal transfer method. As shown in FIG.
and a processing section that electrically processes the signal from the contact sensor section. The contact sensor section was given lecture number 1243 at the Institute of Electronics and Communication Engineers' general national conference in 1981, "High-speed color contact scanner."
As shown in (plate sampling, etc.), a focusing rod lens array (commercially available under the trade name "Selfoc Lens") (17) and COD (19a) to (19e)
) and a light source (not shown). The irradiated light from the light source is reflected on the surface of the document (2υ) and passes through the focusing rod lens array aη to the '411 number (D CCD (L12)
This leads to (19e). The CCDs (L12) to (19e) have a transfer function and serially output signals depending on the amount of received light. In this example, the size of one pixel at the output is determined by CCD (19a) to (19e).
It is read with 6 photosensors. The photo sensor is
Three types of filters, two of the same color, are pasted,
From COD (19a) to (19e), for example, W
,Y. Signals are output in the order c, w, y, c or Y, G, C, Y, G, C (hereinafter repeated). This signal is input to the processing unit (1), and the image forming unit (13)
converted into something suitable for First, the input signal is
The signals are amplified by amplifiers (23a) to (23e) provided corresponding to D (19a) to (19e). next,
The signals are converted into digital signals in AD converters (25a) to (Z5e). This signal is output in time division,
First to third memory CI! 7) , @ , C31
). In other words, are the signals originating from the filters of the same color the same in memory C'? ), 01, C30
are stored sequentially. It is preferable that the first to third memories have a capacity of two scanning lines for reading a document 0υ. Next, the first to third memos 'J (27), CI! !
Signals are sequentially read out from Jl and 03υ and sent to the color processing circuit. This color processing circuit (in 9, c CD
The signals read in Qc+a) to (19e) are converted into signals suitable for the color characteristics of the thermal transfer ink. Details will be described later. Color processing episode Il! The output of the 8-temperature sensor becomes the output of the processing section and is supplied to the image forming section (1, 1). In other words, the ink is transferred to the recording paper by a thermal head from an ink carrier on which each color (K, cyan, magenta (hereinafter abbreviated as M)) and yellow (hereinafter abbreviated as M) is repeatedly applied to each area of an A4 size sheet. Full color is achieved by layering. To realize a color image here, it is necessary to convert the signal by taking into account the color characteristics of the ink and the light receiving characteristics including the filters at COD (19a) to (], 9e). This is realized by special processing circuits. Next, details of the color processing circuit will be explained. The outputs of the first to third memories (27), 01, 01) are the first to third normalization circuits (35a) + (
35h) + (35c). 1st to 3rd
Standardization circuits (35a), (35b), (
35C) has two functions, the first function is a function to remove moiré in reading, and the second function is a function to correct shading and variations in light receiving sensitivity of the photosensor. In order to realize the first function: the first normalization circuit (35
a') is the first memory (
It has a register Oη that holds the signal from 2n for one clock time and an adder G. As a result, two adjacent signals are added. Furthermore, the connection of the output of the adder (3!j) is shifted by 1 bit to the MSB side.Then, the weighted average of two adjacent pixels is taken. Rather than adding pixels, a photosensor is used with filters of the same color pasted on CCDs (19a) to (19e) (the size of a single photosensor is one pixel at the time of input here). This is controlled by a CPU (not shown). Taking the example of reading a monochrome image, we will explain how moiré is removed by averaging the signals in this way. , no filter is pasted on the CCD. First, the number of input (gun) pixels and the number of display pixels are the same, and a sinusoidal image with a period approximately equal to the pitch of the input and display pixels is 1T. Consider the case where an image is input. At this time, since the period of the image and the pitch of the input pixel and the display pixel are the same, the sine wave is sampled at a constant interval T, and a constant value is displayed. However, , when a sinusoidal waveform with a minute difference is input to the pitch T of the input pixel and the display pixel, the (l output value gradually changes) In other words, low frequency (moire occurs).On the other hand, for example, two input pixels (preferably equally spaced) are made to correspond to the display pixel pitch T. When a sinusoidal waveform with a sine waveform is input, the input pixel pitch is half the display pixel pitch T, so the sampling period is approximately half the period of the sine waveform.Then, one input pixel corresponding to the same display pixel and the output signal from the other input pixel is
The signs are opposite and the Zhenfu signals are almost the same. After the external signals are added in this manner, they are supplied to the display pixels as image signals, so no moiré occurs. In addition, the output of the adder (three camphors) is the reading signal of one pixel for each color.The signal processed in accordance with the second function is applied to the signal that has been averaged in this way. The output of the lotus counter C!ω is subtracted in the subtracter (c) using the stored contents of the first coefficient memory (41), and The basics of such hardware are as follows: Now, the white level signal of the standard test pattern is
%V, and the black level signal is 1B, and the output of each sensor is are recorded in the first coefficient memory (30), and the output signals from the adder (30) are used to sequentially input the subtracters (4)
This is performed pixel by pixel. Next 1/(■13
The values of I, v) are recorded in advance in the second coefficient memory (4x), and the multiplier (ID) sequentially multiplies them pixel by pixel. In the same way (ζ1), other color signals are By performing such processing, even if shading occurs in the CC'D sensor, the output signal (signal corresponding to W, Y, C) can be obtained with almost no unevenness in brightness.However, , if there is a change in the characteristics of the color filter, it may be output as a subtle color change.Therefore, next, the first to third coordinate conversion circuits (49a)
1 (49b) and (49c), it is possible to cope with subtle color changes between each photosensor (
As will be described later, the coordinate transformation is performed so that the memory capacity of the color conversion table is small so that the memory capacity of the color conversion table is uniformly distributed in the address space of the color conversion table.The first coordinate conversion circuit ( 49a), as shown in FIG. 4, the first to third normalization circuits (35a), (
35b). Multiplexer Gυ into which the output from (35c) is input
has. The multiplexer r51) includes the first to third normalization circuits (35a), (35b), (3
Switch the signal from 5C) to the product-sum calculation circuit e53)
Output to. In the product-sum operation circuit G:)), the third coefficient memory l55 is
) and performs a predetermined calculation. The purpose of this predetermined operation is to uniformly distribute data in the address space of the color conversion table memory, as described above. That is, in this invention, W of CCD (19a) to (19e). To the Y and C signals, IY and IC send signals corresponding to magenta, yellow, and cyan of the ink ribbon, that is,
The output signals to the image forming unit (131) are obtained.According to the analysis of the present inventor, these output signals are expressed by 1w, IY, xc, and the coefficients are very close to each other. , Teshima C). When you think about it, in the space represented by the variables ■i, IY, and ■c, the output signals are a single group of signals. Then, to accurately represent these output signals, a larger number of bits will be required, and the device ri will become more complex. Therefore, if the color signals are uniformly distributed in the memory space without being distorted all the time, it will be easier to express these signals, and the circuit will also be simpler1. ;i) NoX (19e), 4
For convenience, the output signals from the filters W, Y, and C are referred to as W, Y, and C, and the signals representing the image defined by these signals are expressed as luminance (ij number S1% first and second color difference signals 82. 83 is selected and defined as follows.In other words, 8 may be replaced by W.It is sufficient to define the signal S1 as a signal corresponding to luminance.Equation (1) is Using a matrix, it can be expressed as follows. Next, perform coordinate transformation using the following formula. If the coordinates after O transformation are respectively X+, X2, and X3, the following formula is obtained. Here, the coordinates after transformation are expressed as It is determined to be uniformly distributed on the conversion table address.For example, it is determined to be equidistant for each color of yellow, cyan, and magenta.In this case, the matrix aij in equation (2) is determined by 81 for each color. .Measure S2, 83 (amount of reflected light from the original (
The reflectance) is photoelectrically converted using a photosensor equipped with W, Y, and C filters, and according to equation (1), Sl is 82.83.
I put it on. ), the coordinates X after conversion for that value 1. X
2. I decided on X3. The blank space defined by Xi, X2°X3 may be set as described above at the time of design. Then, by calculating the inverse matrix from these values, atj can be determined.Therefore, even though the same image was read, due to the unevenness of each photosensor itself, 81, 82, 83 for each color is determined for each sensor. Even if they are different, the coordinates XI and X2 after transformation. It is possible to make x3 the same, and the aIJ at that time can be determined. Therefore (1). From (2), the following equation is established. In this way, the signal flow is as follows for the conversion formula (there is only one). In other words, for the standardized w, y, and c signals, the multiplexer 31 selects each of them. However, by multiplying each of them by the coefficient of equation (3) in the product-sum operation circuit Q, Xl, X2, and X3 are obtained. At this time, the coefficients of equation (3) (in the first coordinate conversion circuit (only the coefficients in the first row of 4ω) are determined in advance for each of the seven sensors, and are stored in the third coefficient memory G!5). If this is recorded and calculated by switching for each photosensor, it is possible to eliminate variations between elements. Note that C0D (19a) to (19e
) photosensor, there may be variations from pixel to pixel; in this case, by switching coefficients for each pixel, it is possible to suppress variations in hue from pixel to pixel. If the coefficients aij are determined so that the coordinate-converted addresses are evenly distributed in the coordinate space for each color, the memory capacity required for the color conversion table may be small, as will be described later. In addition, the second and third coordinate conversion circuits (491
) ) , (•19 is a first configuration that is substantially the same as the coordinate conversion circuit (49a), and is different only in the contents of the coefficient memory G5). The signal that has been coordinate-transformed in this way is then subjected to averaging processing, and the color conversion table is drawn using the averaged signal.The local fluctuation rate of the luminance signal is Binarization is performed using a fixed threshold value only when the value is larger than a predetermined value, and dithering is performed in other cases. In other words, in order to express images containing halftones with high fidelity and characters included in the image with high clarity, all images are not expressed using either dither or binary, and image signals are used instead. A beautiful image is displayed by using binary expression in areas where there are severe fluctuations, and by using dither expression in other areas. When Nanakoro reads an image represented by halftone dots using a high-resolution optical system, no matter what pattern the image has, it will be able to read the gaps between the halftone dots. In this case, if you perform dithering and binary switching as described above, the image will be expressed entirely in binary, making the image difficult to view. To explain this situation with concrete numerical values, in the close contact sensor of this embodiment, the image is directly converted into one large satin image by the convergent 11 lens array (17), and the CCD (19a) to (19e)
) is projected to. This is read by CCDs (19a) to (19e). CCD (19a) to (19e,),
The pitch i-,L of the photosensor is about 14 μm,
It is read with a 7z bit photo sensor corresponding to one image line. Furthermore, for the two photosensors that blink at each other, w, y,
A filter of c is attached, and one pixel is read with 6 bits. In other words, 12 lines when separating 3 colors
pm (] pixel is 84 μm), 36 pixels/fight when supporting black and white (
One pixel is approximately μm), and with this resolution, even the gaps between halftone dots can be seen, resulting in so-called halftone noise. Therefore, the filters W, C, Y which are the outputs from CCD (19a) to ff1 (19e) forming the contact sensor
If you look at the rate of change of the signal corresponding to , it is constantly fluctuating. If dither binary values are switched using such a binary signal, an output image is constantly formed in binary values, resulting in an image that is difficult to see. Therefore, instead of the signals corresponding to filters W, C, and Y (signals before conversion), the de-izer binary value is The switching is performed. Furthermore, in this embodiment, the rate of change of the luminance signal is used as the basis for determination, rather than the signals corresponding to W, C, and Y. That is, after performing averaging processing on the luminance signal Xl, the local fluctuation rate is observed. This removes halftone noise, and when dither binary values are switched using this signal, an output image that is easy to see is obtained. For the human eye, a resolution of 2 or 3 lines per line is sufficient for changes in luminance signals. Utilizing this, the above-mentioned averaging process is performed for each 4×4 dot. This averaging process is carried out by the first to third averaging circuits (
57a), (57b), and (57c). That is, the first to third coordinate conversion circuits (5
7a), (57b), and (57c) are sent to the first to third averaging circuits (57a),
Average processing is performed in each of (57b) and (57c). The dither-binary switching described above is a type of edge enhancement. Since the human eye is sensitive to color changes, it is sufficient to use the luminance signal X1 as the standard for dither binary switching. Therefore, the output of the first averaging circuit (57a) is input to the local fluctuation rate measuring circuit 69. This local fluctuation rate measuring circuit 0, as shown in FIG. 5, consists of a subtractor (61). This subtracter (61) takes the difference between the signal that has not been averaged and the signal that has been averaged. This is a differential of a -balance, and the local fluctuation rate can be found with a very simple configuration. To calculate the local fluctuation rate, the first averaging circuit (
It goes without saying that the output of step 57a) may be obtained by performing a distribution process. Using the local fluctuation rate obtained in this way, dither
Before switching the values, a mixing section performs mixing processing on the signal. In other words, the averaged signal and the difference signal representing the local fluctuation rate are combined into a multibreather.
The output is switched and outputted by the trap, and appropriate coefficients are read out from the fourth coefficient memory (to) where appropriate coefficients are stored for both signals, and both signals are mixed at η of the product-sum calculation circuit. . By doing so, it becomes possible to obtain a suitable signal without any blurring. For example, if the signal before averaging is X and the signal after averaging is i, then the subtracter -)
The signals from
) output reaches i. Therefore, if the coefficients of the fourth coefficient memory (2) are a and b, the output y of the product-sum calculation circuit 67 is given by the following equation. y=a(x-ma)+bma (4) Here, if a=b21, then y=X, which is the signal before averaging. If we set a=Q, h:=l, then y-x becomes the averaged signal. Further, if a-=1 and b=o, then y==x-ma, which is a difference signal, that is, a signal with only local fluctuations, in other words, it is close to a differential waveform. This is a low frequency cut signal. By changing the coefficients a and b in this way,
It is possible to freely obtain from the averaged high frequency cut signal to the low frequency cut signal close to the differential waveform. Therefore, by appropriately selecting the coefficients a and I), a signal that is not too blurry can be obtained by taking into account the SN of the signal.
It will be done. For example, when reading an image with a lot of noise, by increasing the coefficient ``s'' and decreasing ``a'', the entire image appears blurred and noise can be suppressed. Also, if you read an image that looks a little blurry, increase a and b
By reducing the size of the image, the image becomes sharper. In this way, it is preferable to change the coefficients a and b depending on the input image, but if this is the case, a corresponding switch or the like may be provided on the operation panel. The other two color difference signals) (2, color conversion unit (69a). (69b) is input to (6!3C). These first to third color conversion units (69a), (69b), (69c
) is constituted by a memory, and the first to third color conversion units (69a) and (69b) are configured based on the signals of these three. (69c) is determined. According to that address, the amount of ink actually needed is retrieved from memory. The color conversion units (69a), (69b), and (69c) are, for example, the well-known Neugebau
It is only necessary to calculate the equation of r in advance and write a memo that constitutes the first to third color conversion sections (69a), (69h), and (69c). Here, we will briefly explain the Neugebauer equation. This equation was thought to be impossible to solve analytically, as explained in, for example, ``Theory of Color Production Eyes'' (written by Neil, published by Printing Society Press). Therefore, the inventors of the present application calculated the Net+gebauer equation as described below, and as shown in FIG. Assume that there are areas (303) and (305) coated with yellow. The area (301) is a white part. Area (:+
O5) is an area where pure colors magenta and yellow are coated overlappingly. Each area (301), (303), (3
05). As can be seen from Figure 19, the plane orientation of (307) is (1
m) (1-Y), (1-m)y, my, m(1-y)
It is. The amount of reflected light when such a unit pixel is read by a photosensor to which a red filter is attached, that is, the reflectance is treated. Furthermore, if the reflectance when reading a pure magenta color using a red filter is expressed as L, and Ry is the case when a pure color of yellow is read, ■=(1-1n)(1-y)・1+
m-(1-y)・Rm+y(1-m)Rym-y-R
It becomes my. Here, assuming that the magenta and yellow inks are transparent, Rmy - 几m, 几)! becomes. Therefore, R=(1-m+mRrn)(1-y+yRy). Similarly, if we consider the case of three colors, we obtain the formula 7. On the other hand, in the embodiment, W, Y, and C are used as filters.
is used, so when converted to the reflectance of these filters, it becomes C'2 (1'G + 13).Now, considering equation (4)7, the signal obtained by the contact sensor is w' , c', y', and the values of J, G, and B. I want to know. However, as seen from (4)', W', C
', Y' alone does not fit R,, G, and B. Therefore, we stop treating it analytically and instead consider all the combinations of variables that satisfy (4)'. For example, when a unit pixel completely covered with yellow is read by a contact sensor, in (4)',
c = man n20, y = 1, and from (4)' we get R, = it, y, o = oy, B by 1) Also, 11. y, ay, BY to W', C', Y
Expressed by ', 几y23 (1-W') -2(1-e') (ly =
2 (]-C')+2 (1-Y')-3(1-W'
)By = 3(] -W') -2(1-Y'). Similarly, Rm, G+n from magenta input
. 13+n, it is possible to determine c, Gc, and Bc, respectively, from the cyan input. Next, in this example, each color is expressed in 16 gradations when printed out, so equation (4)7 for each gradation is expressed as (/
4 Find the value. In other words, specify the gradation for each color, and search for values that satisfy equation (4)7 for all introductions. In the end, the sum of the density of the output image and the read signal obtained from the contact sensor, which are associated with each other, is determined. Therefore, it can be easily solved by recording the relationship between the read signal and the output a "winter degree" in the memory, and determining the output degree from the (1, quiet) signal.On the other hand, local fluctuation rate measurement Output c dimension of circuit (・)!1), 2
Value-dither judgment unit fl) i'i (supplied. This 2
Value-dither judgment unit (7]) irk, 47' in memory
? The external contents are stored in this memory as shown in Table 1. Table 1: Two 's comp
Enter 1 in lemcnt, enter table 4, and input the local support rate as the memory address. -64 to 6;3, which corresponds to the address where the local fluctuation rate of the range 1 river is 1/j, and the dither F J- is the image IZ! in the ζ area. The code that shows the output is carefully noted. l-128 to -
6: When the area corresponding to the address where the local fluctuation rate in the range of 3.64 to 127 is applied is outputted with the value 1-1.2, a code indicating 2 is written (:i'? Therefore, in a small range where the local variable Luri tree is small, the ir:t:, corresponding to the dither, the 1/j code, and when the local variation rate is in the small range, the binary 4Ctl, corresponding to the ir:t:, - code, Each binary value -
Output from the dither judgment section (11)! 11. 1st to 3rd
Image A, K number setting circuit (7″(a), (73b)
, (73c'). These image signal setting circuits (73a), (73b), and (73c) base the output of the binary dither determination unit (T1) on the first to third
The signals of the color conversion units (69a), (69b), and (69C) are converted into signals to be supplied to the surf and throat (not shown). The first Ryōzumi signal setting circuit (73a) converts the signal from the first color converter (69a) into two as shown in FIG.
Includes a binarization circuit (7!5) that converts into values. Here, I will explain the difference between dither expression and binary expression (C) just to be sure. Binary expression corresponds to seven pixels. Comparison with the threshold value is performed, and it is expressed as "0" or "month."On the other hand, dither expression is basically a binary expression, but it corresponds to the density of each predetermined area when reading. Prepare a reference pattern (called a Koreke dither pattern).This reference pattern 1
size, independent output skill (rather than being set up against
The output of the initial number is set as 1. For example 4×
If 4 reference patterns are used, 1 of this reference pattern
For 6 pixels, the priority order is 4 as shown in Figure 8.
The moon is destroyed. That is, the concentration is low.

[Density expression is performed by filling in pixels in order, starting from the lowest number according to the order. Therefore, once the density of a certain region has been detected or defined, a reference pattern is specified according to this density. This reference pattern determines the threshold value set for each pixel. This threshold value cannot be used for binary conversion at 1 m. In this embodiment, in response to a 1 degree change in the reading image gξ,
Dither reference data memory (77), fixed threshold memo January 7
Choose moan. As described above, when the luminance signal is smaller than a predetermined rate of change, binarization is performed using a dither pattern, and in areas where the luminance 41 changes significantly, binarization is performed using a fixed threshold G. This switching is done by the multiplexer (81) to which the output signal from the binary/dither judgment unit (7]) is fed. Switch to
The threshold value data in the binarization circuit C15) is taken in. The binarization circuit (lω) uses this switched threshold value to binarize the signal from the color conversion unit (69a) where i+<1. It is supplied to the thermal head of the thermal printer that constitutes the printer.Furthermore, in areas where the luminance signal changes greatly, such as character patterns, fixed binarization allows the formation of high-resolution, sharply textured images. In addition, for halftone image areas, beautiful halftones can be expressed by using dither patterns.Next, we will explain how to generate standard signals for standardizing image signals (
Next, I'll clarify /Cf61. First, before scanning an input document (91), a standard black and white pattern (93) having a main scanning pattern provided at a scanning start point as shown in FIG. 11 is scanned. As shown in Fig. 12, the interface (
97) to the first to third normalization circuits (35a). (35b), first related meta memo in (35) January 4'D
and 1 to the second coefficient memo. The third coefficient memory 6
Send the following data to: Then, equation (2) or equation (3) becomes as follows.The input data itself is outputted to the output of the product-sum calculation circuit (53). There, the data is input to the CPU (%) via the interface (101). First, enter the standard white pattern. At this time, a slight change in the standard pattern (1δh) is possible, so move the sensor! + Input data several times and send the averaged value to the first coefficient memo (Il) via the interface (97). Next, move the sensor+li
, +17, input the black pattern. At this time, the sensor is moved, and the data is input several times and averaged. Therefore, the CPU (95) divides 1/(IB-I51.) by 1' and sends this value to the second coefficient memory (1') via the interface (97).This value determines the output of each sensor. Standardization is performed. The data necessary for this standardization is obtained just before scanning the original (91), so the light source is stabilized only during the time it takes to scan the original (91). If you set it to Nagisa 1/iL, you can use it even with things that change over time, such as luminescent lamps.Next, I will explain how to determine the coefficients stored in the third coefficient memo υG!'9.Generally There is not much change in the spectral % characteristics of the light source, but depending on the light source, the spectral characteristics of some light sources change over time.
It is necessary to correct the coefficients in this memory 0 to always obtain stable color output. The standard color par is cyan (201
), magenta (203), and yellow rJ (205). You may also use colors such as red, green, and blue.
Further, all of these colors may be used. This standard color par is set as a document, and the coefficients expressed by equation (4) are sent from the CPU (95) to six third coefficient memories via the interface (99). Next, this standard pattern is subjected to shading correction and then read. At this time, the sensor is moved and the data is input several times and averaged, taking into account the contamination of the standard pattern. Do this for each of the three colors. Next, averaging is performed for each sensor, and the coefficients of equation (2) are calculated as described above. Apply this coefficient to the interface (
99) to the third coefficient memory G9. Based on this, it was originally ii;! Even if the color filter of the sensor changes over time, it will correct the output of each -7-7. It will be explained in further detail. The color filters used in the CCD sensor are white, yellow, and cyan, and the output signals from these are w, y, and c, respectively. ). -! : 81 The first color difference is S2, and the color difference can number of RI n2 is S3, and each is expressed as shown in the following equation. As an external luminance signal (rl I = VJ may also be used. Next, in order to prevent a change in luminance even if one hue signal changes, coordinate transformation is performed using the following formula. The coordinates after the transformation Let Xl, X2, and , so that they are equidistant.At this time, the introduction of yellow (output 81. when reading the color).
82.83 are respectively designated as SIY, S2Y, and S3Y, and in the case of cyan, SIC, S2C, and S3C, and further, in the case of magenta, S]M, 82M, and 83M. Also, the coordinates X1 after conversion. XsY when X2 and X3 are yellow. X2Y, X3Y, Shia YTV Toki XxC, X2C
, X3C, 11L Sarah K 7 Zenta time XIM, X2
M,
is the following formula. In this way, even if the address of each color in the color difference space shown by equation (5) differs depending on the image sensor, if the coordinates are transformed using the coefficients shown by equation (8), the addresses can be made the same. becomes possible. Note that the coordinate transformation matrix A does not need to be in the form shown by equation (5), and may be in the form shown by the following equation. In this case, it becomes possible to perform coordinate transformation also in the direction of the luminance signal, and it also becomes possible to perform transformation without deviation in degrees. The method for determining the grandchild's forgiveness is the same as in equation (5) by using the line ;ξ. Now, from equations (6) and (7), X+, , (121'>',': Coefficient a(
If j is multiplied by 4tf in the product-sum operation 9 circuits 6 and 4 and added, the formula <+2 is written. At this time, if the coefficients of the zero-near formula are set in advance for each image sensor, stored in the coefficient memory (), and changed for each image sensor to obtain a total of 1'4, it is possible to It becomes possible to eliminate variations. Note that in a CCD sensor, there may be cases where there is a coefficient of 0-υ for each pixel. In this case, it is possible to suppress variations in hue by changing the coefficient for each pixel. Here, the above β If we define the standard transformation qualitatively, w, c,
If the color of one song i image is expressed by the signal y, then w, c, y
Many colors are concentrated in a very small area of the signal. For example, when color is expressed spatially using W, C, and Y signals (or their difference signals) as independent variables, the w, c, and y coordinate axes ((on the other hand, there are many colors in a small solid) The colors are concentrated, and the color changes drastically in response to minute changes in w, c, and y. Expressing such colors using w, c, and y requires a large number of bits, which is not a good idea. Therefore, it is preferable to change the coordinate axes to expand the space in which colors are concentrated.The above-mentioned coordinate transformation achieves this.After further study by the present inventors, the coordinate transformation is performed, Even if the signals are distributed uniformly in the memory space, it is difficult to express high-density colors, and it is not possible to distinguish between signals with the same number of bits as in other areas, and all high-density colors are Therefore, a non-linear transformation is applied to the luminance signal, the color difference signal, and the rib luminance signal. In other words, as shown in Fig. 15, the mixing section (6:3)
The output of the product-sum arithmetic circuit (67), which is the output of The output of the mixer (63) is supplied to this memory (207) as its address. Jin's memory (2)
07) is a logarithmic function, as shown in FIG. 14, and is set so that the output increases more significantly as the input address becomes larger. Mixing unit (63) given as address as described above
The output of the output (l) increases as the value increases, and the difference between the output values increases, but by logarithmically converting the interval between the luminance signals, the interval between the luminance signals becomes narrower. 0 The contact sensor in this embodiment has a very high resolution.CCD (19a) to C19e
) As shown in FIG. 16(a), each of the seven photosensors has a single color filter attached to it, for example, three types of filters w, y, and c. One pixel is read by three 7-photosensors provided with the same color (or the same is true for three pairs of photosensors provided with filters of the same color). This is called a space-divided color sensor. In case (B), the color of the document suddenly changes in the middle of the filter group corresponding to one pixel, as shown in (b) of the same figure. Suppose then. Then, the W filter reads the pure magenta color, and the signal W(M) becomes Y. C filter reads pure cyan color, Y(C)
. A signal C (C) is output from the CCD. In principle, one color is specified by the output signal of a photosensor to which three filters, IiW;'Y, and C, are attached. However, in this case, W(8i), y(
The color specified by the combination of signals c) and c(c) does not exist. This means that it overflows in the memory space. Such a situation can occur, for example, at the edges of characters. This situation also occurs when the size of halftone dots forming an image is approximately the size of a photosensor. In the former case, even if a non-existent color combination occurs in the green of the letter, 1-1] to 2 will occur even for 1T17i element of Only. However, in the case of scoundrels, ]Ii! When the above-mentioned situation occurs with moisture, it often occurs in the entire nearby area, and a large number of non-existent color signals are combined for one scanning line. Put it away. Therefore, the prescribed reading J
If a combination of more than a predetermined number of non-existent color signals occurs in the r9 region, what happens? 3,? It is determined to be noise,
More averaging over the signal. This reduces halftone noise. This embodiment is shown in FIG. In the apparatus according to the embodiment shown in FIG. 2, the first to third averaging circuits (57a
), (57b), (57C) The circuit configuration between the output terminals and the input terminal of the image forming section 03) is distinctive. A signal X1 corresponding to the luminance output from the mixing unit (63),
The noise detection unit (2) uses the color difference signals X2 and
09), fine noise is detected. According to this detection result, the first to third average mouth #S (57a), (5
Using either of the output signals of 7b) and (57c) and the further averaged signal, a signal corresponding to the amount of ink transferred in the image forming section (1) is determined. ), the outputs of the second and third averaging circuits (571)), and (57c) are sent to the color judgment circuit (21
.
The memory 2 that constitutes the memory 2 is composed of memories having the same address space, but the contents thereof are different. That is, when a color consisting of a combination of three output signals exists, Ill,
If it does not exist, "0" is stored. As mentioned above, when the contact sensor of this embodiment discards halftone noise, the output of the color determination circuit (211) becomes [month]. A counter circuit (213) counts this number within a predetermined area. This count value and the threshold value set by the threshold value circuit (215) are compared in the determination circuit (217), and if it is larger than the threshold value, for example, "1" is output. This is a signal indicating that halftone noise has occurred. On the other hand, the second and If', 3 average circuits (574)),
The output signal of (57c) is the averaging circuit (219), (2
21), each average is taken. Here, the averaging process in the first to third averaging circuits (57a), (57b), and (57C) and the averaging circuit (219). The average processing in (221) and the width of the nitrogen range to be processed are different. In the former case, an average is taken for 2×2 pixels. In the latter case, only 2×2 signals of these averaged signals are averaged. Therefore, the latter averaging is performed over a wider space. This (qJl, like the former, X11 point noise is the second
And since the average processing in the third averaging circuits (57b) and (57c) does not remove them, further averaging processing is performed. Therefore, the output of the judgment circuit (2]7) is used to send these two signals to the multiplexers (223) and (225).
Switch with . Furthermore, for the luminance signal, the outputs (a, b) of the mixer (63) are sent to the multiplexer (227) K.
The characteristics of the signal change depending on the value of , but here it is an averaged signal. ) and the first averaging circuit (57a)
The output of the determination circuit (217) is switched depending on the output of the determination circuit (217). In this way, the first to third color conversion units (69a), (69
1, +). (69c) is supplied with a signal that is significantly affected by noise at points Id and W. These color conversion units (69a), (69h),
(69c) is the same as in the previous embodiment, and determines the amount of ink required. In this way, the color conversion section (69; 1)
, (691)) and (69C), the binarization circuit (7
5a), (75b), and (75c). The binarization circuits (75a), (75b), and (75C) are also the same as those in the actual tjni example described above, and the dither reference data memories (77a) and (7711). (77c), fixed threshold memory (79a), (79t)
), (79c) to the multiplexer (81a),
(81b) and (81c) are supplied to the switching circuits (75a), (75b), and (75c). Binarization circuit (75a), (75b), (75c)-
is a comparison circuit, q is a multiplexer (81a), (8
1b), (81c) The signals supplied from KA (bm, t-c) are used as reference values for comparison. ? Lutiplexer (81a), (81b), (81c)
The switching is. As described above, switching is performed by the output of the binary/dither determination circuit qO. As mentioned above, this is a method in which, when characters, etc. are mixed in a halftone image, the characters, etc. are displayed with rounded edges. Furthermore, as is clear from the explanation up to this point, if single-point noise occurs, regardless of whether the luminance signal fluctuates sharply or not, it is necessary to dither the signal, which has been averaged to a large degree with respect to the pixel size. Expression 1 is good. Therefore, the determination circuit (217)
If the output of [month] and the occurrence of network noise is detected, the multiplexers (81a), (81b), (81c)
Therefore, the dither reference data memory (79a),
Select (79b) and (79c). Figure 18 shows the input image, and targets normal printed matter.
This is an example of simplification of the circuit. First, let's consider that the input drawings consist of binary images of d characters and reduced-scale images, and halftone images of halftone dot patterns. Therefore, if it is a halftone dot pattern, the signal is averaged and dithered. If the signal is not narrow or patterned, the signal is simply binarized and output without averaging. That is, in FIG. 18, the nII signal to be averaged is inputted to the noise detection section (209), and the output of the determination circuit (217) is used as the signal of the fourth to sixth averaging circuits (251a), (251).
The outputs of h), (251c) or the signals before averaging are sent to multiplexers (223), (225), (22
Select in 7). The output of the lever judgment circuit 32 selects the dither pattern 27 or the fixed value part. Note that this circuit uses a common dither pattern for each color. By doing this, the influence of noise caused by halftone dots can be reduced. The half-tone display becomes OK11 without being affected. Also, line drawings such as characters can be output at high resolution. [Other Embodiments of the Invention] In the above embodiments, yellow, magenta Although the description has been made for three colors using a cyan ink ribbon, the same process can be performed for four colors using a black ink ribbon.In other words, as shown in FIG.
color conversion section (69d), fourth image signal setting circuit (73
d). The rest is the same as the previous embodiment. At this time, if the signals from the W, Y, and C filters of CCD (19a) to (19e 'I) are all [moon (a predetermined value or more), black is used instead of mixing the three colors. In the embodiment described above, the difference signal is used as an input signal for the binary dither judgment unit (7]), but in this case, continuity is lost in the memory contents. Instead, if one bit is added to the MOB and the result is used as the address of the memory constituting the binary dither judgment section (71), continuity will occur in the contents of the memory. If the fixed threshold value is made common to each color, the circuit is omitted as shown in Fig. 10.In addition, in the case of non-destructive variation 1 negative unhealthy case, fqt, , memory (
69a), (69ri), and (69c) may be omitted. However, in the previous embodiment, the image sensor is one-dimensional COD.
Although an example of a sensor has been described, a two-dimensional sensor may also be used.

[Brief explanation of drawings]

FIG. 1 is a diagram showing a schematic configuration of a color photographic camera in an embodiment, FIG. 2 is a diagram showing details of the color processing circuit shown in FIG. 1, and FIGS. 8 is a diagram for explaining dither expression, FIG. 9 is a diagram showing a color processing circuit when four color signals are used for output, and FIG. A diagram showing an example of fm standardization of the color processing circuit shown in Figure 2, Figures 11 to 13 are diagrams for explaining the standardization of image signals, and Figures 14 and 15 are diagrams showing the logarithm of the image signal. Figure 16 is a diagram to explain the conversion, Figure 16 is a diagram to explain noise to the mesh in a space division type sensor, Figures 17 and 18 are diagrams to effectively reduce halftone noise in a space division type sensor. FIG. 19 is a diagram for explaining the Neugellauer equation. (57a), (57b), (57c) --- First to third averaging circuits, (59 )...Local fluctuation rate measuring circuit, (71)...Binary/dither determination section, (73a), (73b), (73C), . . ., first to third image signal setting circuits. L'---'' s3 Figure 4 Figure 6 Figure 9 Figure 10 Figure 15 Figure 16

Claims (1)

[Claims] (1) An averaging means that averages a plurality of image signals representing image information of a minute area of a document, and a binarization circuit that binarizes the image signal averaged by the averaging means using a threshold value. , comprising a local fluctuation rate measuring means for measuring the fluctuation rate of the image signal obtained by the averaging means, and when the fluctuation rate obtained by the measuring means is larger than a predetermined value, the threshold value is used for fixed binary expression. An image signal processing apparatus characterized in that the first threshold value for dither expression is selected as the threshold value, and if the variation rate is smaller than a predetermined value, the second threshold value for dither expression is selected as the threshold value.