JPH09265510A - Optical character reader - Google Patents

Abstract

PROBLEM TO BE SOLVED: To provide an OCR(optical character reader) which makes an accurate color decision without being affected by the relative output level difference of a color image sensor and surely removes dropout colors. SOLUTION: Color signals rd, gr, and b1 of red, green, and blue are supplied to color decision parts 31-34 and a maximum detection part 35. The color decision parts 31-34 make color decisions on the basis of reference values of mutually different classes. The maximum value detection part 35 detects a maximum value max among the color signals and a peak detection part 50 detects a peak value peak from the maximum value max. A level decision part 60 decides the class of a color signal level from the peak value peak and supplies a decision result csel to a color information selection part 40. The color information selection part 40 selects a color decision result bsel corresponding to the color signal level out of color decision results c11-c14 outputted from the color decision parts 31-34. A luminance signal selection part 90 selects and outputs a luminance signal bw according to the selected color decision result bsel.

Description

Detailed Description of the Invention

[0001]

BACKGROUND OF THE INVENTION 1. Field of the Invention The present invention relates to an optical character reader (hereinafter referred to as "OCR"), and more particularly to processing dropout color in an image image read by a color image sensor.

[0002]

2. Description of the Related Art FIG. 2 is a block diagram showing a schematic configuration of a conventional OCR. The OCR is a device that optically reads characters on the form 1. The form 1 in FIG. 2 is a sheet for entering characters in a certain format and inputting into the OCR, and a character frame 1a for entering characters is printed in advance. Since the character frame 1a is not necessary for reading characters, it is visible to the writer but it is desirable that it cannot be mechanically read. Therefore, the character frame 1a is printed in a chromatic color (dropout color) for mechanically ignoring it. And in this character frame 1a,
The character 1b to be recognized is written in an achromatic color (for example, black). Such a form 1 is optically read by the OCR color image reading unit 2. The color image reading unit 2 is, for example, a charge coupled device (Charge Coupled D).
evice, hereinafter referred to as CCD), and is composed of a line type color image sensor or the like. The color image reading unit 2 decomposes the characters and other information described in the form 1 into pixels and optically reads them, and a red component signal (hereinafter, referred to as R signal) rd and a green component signal (hereinafter, referred to as R signal) for each pixel. , G
And a blue component signal (hereinafter, referred to as B signal) bl. The output of the color image reading unit 2 is given to the color processing unit 3. The color processing unit 3
On the basis of the R, G, B signals rd, gr, bl, pixels of colors other than characters are extracted and converted into the same color as the background color, thereby removing the character frame in the image data. Image data img with the character frame removed in this way
Are stored in the image memory 4. Image memory 4
The cutout control unit 5 cuts out a predetermined area in which the character pattern exists from the image data img stored in. The cutout control unit 5 stores the cut-out character pattern in the character pattern memory 6. The character pattern stored in the character pattern memory 6 is read by the character recognition unit 7 and is recognized by referring to the recognition dictionary 8.

FIG. 3 is a block diagram showing an example of the configuration of a conventional color processing section 3 in the OCR of FIG. The color processing section 3 has a sensor interface section 11 for inputting the R, G, B signals rd, gr, bl output from the color image reading section 2 of FIG. The output side of the sensor interface unit 11 is connected to the RGB alignment unit 12. The RGB alignment unit 12 samples the R, G, and B signals rd, gr, and bl, and corrects the positional deviation of the red, green, and blue pixels due to the pixel array of the color image sensor. G and B signals rd and g
r and bl are aligned at the same position. The output side of the RGB alignment unit 12 has a color determination unit 13 and a maximum value detection unit 1
4 is commonly connected. The color determination unit 13 uses R, G, B
The level difference between the signals rd, gr, and bl is calculated, and the level difference is compared with a specific criterion to determine whether the pixel is chromatic or achromatic. Then, the determination result is used as a color information signal (red color red, green color grn, blue color bl).
u or an achromatic color blk). Further, the maximum value detection unit 14 outputs the highest signal level of the R, G, B signals rd, gr, bl as the maximum value signal max of the pixel. The output side of the color determination unit 13 is connected to the color information processing unit 15. Color information processing unit 15
Is a mode selection signal mo provided from the output switching unit 16.
attribute (character / color) switching signal at based on de
r, a color (chromatic / achromatic color) determination signal col, and a luminance selection signal bsel are output. The luminance selection signal bsel of the color information processing unit 15 corresponds to the luminance signal selection unit 17
Given to. The luminance signal selection unit 17 includes the RGB alignment unit 1
R, G, B signals rd, gr, bl output from 2
Among the maximum value signals max output from the maximum value detection unit 14, any one is selected based on the brightness selection signal bsel and the brightness signal bw is output.

On the other hand, the maximum value signal max is given to the peak detector 18. The peak detector 18 detects the peak of the signal level of the input maximum value signal max, holds it as the peak value peak, and outputs it. The output side of the peak detection section 18 and the output side of the luminance signal selection section 17 are both connected to the luminance signal correction section 19. The brightness signal correction unit 19 uses the peak value peak input from the peak detection unit 18 to calculate the brightness signal bw.
The level of is corrected. This brightness signal correction unit 1
The output side of 9 is connected to the image memory 4 of FIG. In the color processing unit 3 having such a configuration, the following operation is performed. The form 1 is optically read by the color image reading unit 2 in FIG. 2, and characters and other information on the form are separated into pixels and separated into respective primary color components of red, green and blue. Then, R, G, and B signals rd, gr, and bl are input to the RGB alignment unit 12 via the sensor interface 11. In the RGB alignment unit 12, the input R,
The signals rd, gr and bl of G and B are shown in FIG.
The following processing is performed according to the pixel array of the image sensor of the color image reading unit 2 shown in (4).

4 (1) to 4 (4) are diagrams showing an example of a pixel array pattern of a line type color image sensor composed of a CCD. For example, the RGB alignment unit 12 using the image sensor of the array of FIG. 4A has a delay register circuit for correcting the positional deviation of each pixel of red, green and blue. The RGB alignment unit 12 using the image sensors of the arrangements of (2) and (3) of FIG. 4 has a function of aligning R, G, and B signals. The image sensor of the array of FIG. 4 (4) does not have the positional deviation of each pixel of red, green, and blue, and thus no special processing is required. After the processing according to the pixel array of such an image sensor is performed, R, G, B signals rd, g
r and bl are input to the color determination unit 13. Color determination unit 13
Then, the level difference between the R, G, and B signals rd, gr, bl is calculated, and the signal having the highest level and having a level difference equal to or higher than a specific determination reference value with respect to other signals is calculated. select. Then, it is determined that the color of the selected signal is the color of the pixel on the form 1. If there is no signal having a level difference equal to or more than the specific determination reference value, it is determined that the color of the pixel on the form 1 is an achromatic color. As a result, the colors on the form 1 can be identified, and if the color is achromatic, it is determined that the data is image data forming a character.

[0006]

However, the conventional OCR has the following problems. The color determination unit 13 in FIG. 3 calculates a difference in signal level for each combination of the R signal rd and the G signal gr, the G signal gr and the B signal b, and the B signal bl and the R signal rd. Then, when any one of the level differences is equal to or larger than the specific determination reference value, the color of the signal having the highest level is determined to be the color of the pixel. For example, the signal level of the R signal rd is the G signal gr.
And when the signal level of the R signal rd is higher than the signal level of the B signal bl by a certain determination reference value or more than the signal level of the G signal gr, the pixel is determined to be a reddish color. When it is determined that the color is red, only the red color information signal red in the output signal of the color determination unit 13 is turned on and all other signals are turned off. R signal rd, G signal gr,
If there is no difference between the B signal and the B signal bl that is greater than or equal to the specific determination reference value, the pixel is determined to be achromatic and the achromatic color information signal blk is turned on. Since the color determination unit 13 uses such a color determination method, when the output levels of the color image sensors are relatively different, the color determination method using the level difference of the specific determination reference value is as follows. Figure 5 (a)
It becomes impossible for the reason shown in (c).

FIGS. 5A to 5C are explanatory diagrams for explaining the problems of the conventional color determination method. FIG. 5 (a)
Indicates a state in which the color sensor is reading the reddish color portion of the form 1 when red is used as the dropout color, for example. FIG. 5B shows the signal levels of the R signal rd, the G signal gr, and the B signal bl output from the color sensor A. Here, the output level is 256 gradations, the R signal rd, the G signal gr, and the B signal b.
The signal levels of l are 200, 140 and 140, respectively.
It is. When the level difference ΔL of the determination standard is 50, the level difference ΔRG between the R signal and the G signal is ΔRG = R signal level−G signal level = 60> 50. That is, it is determined that the color is a dropout color.

On the other hand, FIG. 5C shows the R signal rd, the G signal gr, and the B signal bl output from the color sensor B. The signal levels of the R signal rd, the G signal gr, and the B signal bl are , 140, 100, 100, respectively.
Therefore, the level difference ΔRG between the R signal and the G signal and the level difference ΔRB between the R signal and the B signal are respectively ΔRG = R signal level−G signal level = 40 <50 ΔRB = R signal level−B signal level = 40 < Therefore, the color determination unit 13 determines that the color is achromatic. In this way, the R signal r output from the color sensor B
Even if the relative level relationship between the d signal, the G signal gr, and the B signal bl is the same as that of the color sensor A, the output itself is relatively low, so the level difference between the signal levels becomes small, and the specific judgment reference value If the level difference is determined at (50 in this case), different determination results will occur. The present invention solves the problems of the above-mentioned conventional techniques, corrects the color judgment level for the relative level difference of the output signal level of the image sensor, performs accurate color processing, and ensures dropout color. To provide OCR for removal.

[0009]

In order to solve the above-mentioned problems, the first invention is to decompose a character described on a form into pixels and read them optically to obtain primary color component signals of red, green and blue. Color image reading means that outputs separately, and color processing that performs color determination based on the signal level difference between each of the primary color component signals and outputs the determined red, green, blue or achromatic color component signal as image data. In the optical character reading device, the color processing includes: means for recognizing a character written on the form by referring to a registered pattern registered in a character recognition dictionary from the image data. The means is configured as follows. That is, the color processing means performs color determination using different reference values for the signal level differences, and outputs a plurality of determination results as color information signals, and the primary color component signals. The maximum value detection unit that detects and outputs the maximum value signal having the maximum signal level from among them, and the color information signal output from the plurality of color determination units is selected by the color information selection signal. A color information selection unit for outputting as a signal. Further, the color processing unit detects a maximum value of the signal level of the maximum value signal, holds a peak value and outputs the peak value, and determines a class of the signal level difference based on the peak value. A level determination unit that outputs the color information selection signal according to a determination result; a color information processing unit that outputs a luminance selection signal based on the color signal and a mode selection signal given according to a processing mode; A brightness signal selection unit that selects one of the primary color component signals and the maximum value signal based on a brightness selection signal and outputs the brightness signal, and a signal level of the brightness signal is corrected based on the peak value. And a luminance signal correction section for outputting image data to the recognition means.

According to a second aspect of the present invention, in the color processing means of the first aspect, a second peak detecting section for detecting the peak of the signal level of the luminance signal and holding and outputting it as a luminance correction signal is added. There is. Then, the brightness signal correction unit, based on the brightness correction signal instead of the peak value,
The signal level of the luminance signal is corrected so that the image data is output to the recognition means. In the third invention, instead of the color information selecting section, the first peak detecting section and the level determining section of the color processing means of the second invention, the following first peak detecting section and peak having different functions are provided. A maximum value output unit, a brightness level selection unit, and a level determination unit are provided. The first peak detector of the third invention is the red detector,
A signal level peak is detected for each of the green and blue primary color component signals, and the peaks are held and output as peak values. The peak maximum value output unit outputs the maximum value among the plurality of peak values output from the first peak detection unit as the maximum peak value. The brightness level selection unit selects one of a plurality of peak values output from the first peak detection unit and the maximum peak value by a mode selection signal given according to a processing mode, and selects a brightness level value. Is output as. The level determination unit determines the class of the signal level difference based on the brightness level value and outputs the color information selection signal according to the determination result.

According to a fourth aspect of the invention, an interpolation signal generating section having the following function is added to the color processing means of the third aspect. The interpolation signal generation unit of the fourth invention interpolates the primary color component signal at a pixel position where the primary color component signal does not exist from the signal levels of the primary color component signals of the same color on both sides of the pixel position to interpolate the primary color component signal. Is generated. With the addition of the interpolation signal generation unit, in the fourth invention, the functions of the first peak detection unit and the luminance signal selection unit are changed as follows. That is, the first peak detector detects the peak of the signal level for each of the red, green, and blue primary color component signals and the interpolated primary color component signals, and holds and outputs each as a peak value. Further, the luminance signal selection unit selects, based on the luminance selection signal, one of the primary color component signals or one of the primary color component signals and the interpolated primary color component signals and outputs the selected luminance signal. I have to. In a fifth invention, in the first, second, third or fourth invention,
The number of classes of the signal level difference determined by the level determination unit is the same as the number of the plurality of color determination units, and the level determination unit is responsive to each reference value used by the plurality of color determination units for color determination. Then, the class is judged.

According to the first invention, the OCR is performed as described above.
As described above, the OCR color processing means performs the following operations. From the color image reading means red,
When the green and blue primary color component signals are input, the plurality of color determination sections perform color determination using different reference values, and output color information signals, respectively. Also, the maximum value detection unit
The maximum signal level of the input primary color component signals of red, green, and blue is detected and the maximum value signal is output. The peak detection unit detects the peak of the maximum value signal and outputs the detected peak value to the level determination unit. The level determination unit determines a class of reference values used for color determination based on the peak value and outputs a color information selection signal. The color information selection unit selects one of the color information signals input from the plurality of color determination units based on the color information selection signal and outputs the color signal. The color information processing unit outputs a luminance selection signal based on the color signal and the mode selection signal. The brightness signal selection unit selects one of the primary color component signals of red, green, and blue and the maximum value signal based on the brightness selection signal, and outputs the brightness signal. The brightness signal correction unit corrects the signal level of the brightness signal based on the peak value and outputs it as image data to the recognition means. In the second invention, as in the first invention, the luminance signal selection section outputs the luminance signal. The output brightness signal is provided to the second peak detection unit and the brightness signal correction unit. The second peak detection unit detects the peak of the input luminance signal and outputs the luminance correction signal. The brightness signal correction unit corrects the signal level of the brightness signal based on the brightness correction signal and outputs it as image data to the recognition means.

In the third aspect of the invention, when the red, green, and blue primary color component signals are input from the color image reading means, the first peak detection section outputs signals for each of the red, green, and blue primary color component signals. The peak of the level is detected and each peak value is output. The peak maximum value output unit outputs the maximum value among the red, green, and blue peak values as the maximum peak value. The brightness level selection unit selects one of the red, green, and blue peak values and the maximum peak value by the mode selection signal and outputs the brightness level value. The level determination unit determines a class of reference values used for color determination based on the brightness level value and outputs a color information selection signal. The color information selection unit selects one of the color information signals input from the plurality of color determination units based on the color information selection signal and outputs the luminance selection signal. The processing for the luminance signal output from the luminance signal selection unit based on the luminance selection signal is the same as that of the second invention. In the fourth invention, when the red, green, and blue primary color component signals are input from the color image reading means, the interpolation signal generating section of the color processing means outputs the primary color component signals at pixel positions where no primary color component signal exists. Interpolation primary color component signals are generated by interpolating from the signal levels of the primary color component signals of the same color on both sides of the pixel position. Then, based on the generated interpolated primary color component signals and the input primary color component signals of red, green, and blue, color processing substantially similar to that of the third invention is performed.

[0014]

First Embodiment FIG. 1 shows a first embodiment of the present invention. For example, a color processing unit provided in place of the color processing unit 3 in the OCR of FIG. It is a block diagram which shows a structure. The color processing section has a sensor interface section 21 for inputting R, G, B signals rd, gr, bl output from the color image reading section 2 of FIG. 2, for example. The output side of the sensor interface unit 21 is connected to the RGB alignment unit 22. The RGB alignment unit 22 samples the R, G, B signals rd, gr, bl. And FIG.
Of the red, green, and blue pixels due to the pixel array of the color image sensor in the color image reading unit 2 of FIG. 2 is corrected, and the R, G, and B signals rd, gr, and bl are aligned at the same position. It is what makes me. The output side of the RGB aligning unit 22 is connected to the color input unit 30 by, for example, eight signal lines for displaying 256 gradation levels. The color input unit 30 includes color determination units 31, 32, 33,
34 and a maximum value detection unit 35. The color determination units 31 to 34 respectively output the R, G, and B signals rd,
By calculating a level difference between gr and bl and comparing the level differences using different reference values for each color determination unit, it is possible to determine whether the pixel is a reddish color, a greenish color, or a bluish color. Or, determine whether it is an achromatic color. Then, the determination results are output as color information signals cl1, cl2, cl3, cl4, respectively. The output side of each of the color determination sections 31 to 34 is connected to the color information selection section 40 by two signal lines for displaying four types of color information. On the other hand, the maximum value detection unit 35 outputs the highest signal level of the R, G, B signals rd, gr, bl as the maximum value signal max of the pixel. The output side of the maximum value detection unit 35 is connected to the peak detection unit 50.

The peak detector 50 has a memory for storing the maximum value of the level of the input maximum value signal max, and updates the stored content when a level higher than the stored level is input. . In this way, the peak of the input maximum value signal max is held as the peak value peak and output. The output side of the peak detection unit 50 is connected to the level determination unit 60. The level determination unit 60 determines the peak value peak from the four color determination units 31 to 31.
Divided into 4 classes corresponding to 34 detection levels,
The result of the classification is output as the color information selection signal csel. The output side of the level determination unit 60 is connected to the selection terminal SEL of the color information signal selection unit 40. The color information selection unit 40 receives the color information signals cl1 to cl input from the color determination units 31 to 34 based on the color information selection signal csel.
One of the four signals is selected and the color signal col is output, and its output side is connected to the color information processing unit 70. The color information processing unit 70, based on the mode selection signal mode given from the output switching unit 80, attributes (character / character / character).
A color) switching signal art, a color (chromatic / achromatic color) determination signal col, and a luminance selection signal bsel are output. The luminance selection signal bsel output from the color information processing unit 70 is given to the luminance signal selection unit 90. The luminance signal selection unit 90 includes R, G, and B signals rd, gr, bl output from the RGB alignment unit 22 and the maximum value detection unit 3.
One of the maximum value signals max output from 5 is selected based on the brightness selection signal bsel and output as the brightness signal bw. The output side of the brightness signal selection unit 90 and the output side of the peak detection unit 50 are connected to the input side of the brightness signal correction unit 100, respectively. The brightness signal correction unit 100 uses the peak value p supplied from the peak detection unit 50.
The eak corrects the signal level of the luminance signal bw. The output side of the luminance signal correction unit 100 is connected to the dropout processing unit 110. The dropout processing unit 110, based on the attribute switching signal atr and the color determination signal col output from the color information processing unit 70,
The dropout color is removed by determining whether or not the pixel is a dropout color. The output side of the dropout processing unit 110 is the image memory 4 of FIG.
It is connected to the.

FIG. 6 is a block diagram showing the configuration of the peak detector 50 of FIG. The peak detection unit 50 has a peak data comparison unit 51 to which the maximum value signal max is input. The peak data comparison unit 51 receives the maximum value signal max and the random access memory (hereinafter referred to as RAM).
The data held in 52) is compared and the update signal ren is output when the maximum value signal max is larger than the data in the RAM 52. The output side of the peak data comparison unit 51 has a peak data update determination unit 53.
Connected to The peak data update determination unit 53 only operates when the peak reset signal rst is given.
According to the input update signal ren, the data switching signal c
ng is turned on for output. The output side of the peak data update determination unit 53 is connected to the control terminal C of the update data switching unit 54. The update data switching unit 54 has a terminal A to which the maximum value signal max is input and a terminal B to which the data held in the RAM 52 is input, and is input to the terminal A while the data switching signal cng is on. The output signal is output to the output terminal O. The output terminal O of the update data switching unit 54 is the data input terminal D of the RAM 52.
I. The RAM 52 reads / writes the data of the address according to the address signal adr provided from the address control unit 55 connected to the address terminal A, and has a capacity of the number of pixels corresponding to one line of the line type color image sensor. Have The output of the update data switching unit 54 is RA as the peak value peak.
The data is written and stored in M52, and is output to the level determination unit 60 and the luminance signal correction unit 100 in FIG.

Next, the operation of the color processing section of FIG. 1 will be described. First, the R, G, B signals rd, gr, bl aligned by the RGB alignment unit 22 are input to the color determination units 31-34. Each of the color determination units 31 to 34 individually has a reference value (for example, a detection parameter) for performing color determination, and the detection parameters and the input R, G, B signals rd, gr, b.
and the level difference between l and the result is compared with the color information signal cl
Output as 1 to cl4. Here, each color determination unit 31 to
The judgment criteria of 34 are classes 1, 2, 3, and 4, respectively. The detection parameters for each class are set as follows, for example. The color determination unit 41 of class 1 uses R, G,
The detection parameters for the B signals rd, gr, and bl are ΔR1, ΔG1, and ΔB1, respectively. The class 2 color determination unit 32 determines that the output level of the color determination unit 31 is uniformly reduced by 10% irrespective of the color difference, and the detection parameters for the R, G, B signals rd, gr, bl are each ΔR1. X0.9, ΔG1 × 0.9, ΔB1 × 0.9. The class 3 color determination unit 33 determines that the output level of the color determination unit 31 is uniformly reduced by 20% irrespective of the color difference, and the detection parameters for the R, G, B signals rd, gr, bl are each ΔR1. × 0.8, ΔG1 × 0.
8, ΔB1 × 0.8. Then, the class 4 color determination unit 34 determines that the output level of the color determination unit 31 is uniformly reduced by 30% irrespective of the color difference, and detects the detection parameters for the R, G, B signals rd, gr, bl. Let ΔR1 × 0.7, ΔG1 × 0.7, and ΔB1 × 0.7, respectively. Thus, for example, when the detection parameter of class 1 is ΔR1 = 50, the color information signals cl1 to cl4 output by the color determination units 31 to 34 at each level when the sensor A of FIG. 5 is used are all red. A signal indicating a system color is used. On the other hand, when sensor B is used, classes 2, 3, 4
The detection parameters of ΔR2 = 45, ΔR3 = 40,
Since ΔR4 = 35, the color determination units 32, 33, and 34 of classes 2, 3, and 4 have color information signals cl2 to cl2 indicating a reddish color.
cl4 is output, and the class 1 color determination unit 31 outputs a color information signal cl1 indicating an achromatic color. The outputs of these color determination units 31 to 34 are given to the color information signal selection unit 40.

On the other hand, the maximum value detecting section 35 detects the maximum value signal max from the R, G, B signals rd, gr, bl as the maximum value signal max and outputs it to the peak detecting section 50. Here, the operation of the peak detector 50 will be described with reference to FIG. The maximum value signal max is input to the peak data comparison unit 51 in FIG. Peak data comparison unit 5
The data from the RAM 52 is further input to 1. Now, it is assumed that the maximum value signal max is larger than the data input from the RAM 52. Since the update signal ren output from the peak data comparison unit 51 is turned on, the peak data update determination unit 53 outputs the data switching signal cng to the control terminal C of the update data switching unit 54. When the data switching signal cng is input to the control terminal C, the update data switching unit 54 receives the maximum value signal ma input to the terminal A.
x is output to the output terminal O. The output terminal O is the RAM 52
RAM5 because it is connected to the data input terminal DI of
The data of 4 is rewritten to the maximum value signal max. R
Since the address corresponding to the pixel being compared by the peak data comparison unit 51 is input from the address control unit 55 to the address terminal A of the AM 52, the maximum value signal max for each pixel in one line is updated and the peak is obtained. The value peak
Is stored in the RAM 52 as In such a peak value updating process, immediately after the reading of the form is started, the peak reset signal rs is spread over a plurality of lines corresponding to the margin portion at the leading end.
This is done by turning on t. An example of this processing is shown in FIG.

FIG. 7 is a diagram showing an example of the relationship between the reading position of the form and the peak reset signal rst. When the peak reset signal rst is turned off, the peak value update processing is not performed, and the data stored and held in the RAM 52 is output from the update data switching unit 54 as the peak value peak. In this way, the peak detection unit 50 outputs the maximum signal level of the input maximum value signals max to the level determination unit 60 as the peak value peak. The level determination unit 60 classifies the output levels of the color image sensor based on the input peak value peak as shown in FIG. FIG. 8 is an explanatory diagram showing an example of processing in the level determination unit 60. As shown in FIG. 8, the peak value peak is used as a standard of the output level of the color image sensor to divide the signal power level into four stages. That is, 100 to 90% for class 1 and 90 for class 2.
~ 80%, Class 3 80-70%, Class 4 70%
Below, the judgment level of each class is set. The level determination unit 60 compares these determination levels with the input peak value peak and outputs a signal indicating the corresponding class as the color information selection signal csel. For example, when the peak value peak of the sensor A in FIG. 5 is the level value 240, the level determination unit 60 outputs class 1 as the color information selection signal csel, and the peak value peak of the sensor B is the level value 160. And class 4 is output. Then, the color information selection signal csel as the determination result is output to the color information signal selection unit 40. In the case of the sensor B, since it is determined to be class 4, the color information signal selection unit 40 selects and outputs the color determination unit of class 4, that is, the color information signal cl4 of the color determination unit 34. Therefore, since the detection result of the class 4 of the sensor B shows a reddish color, the color recognition result of the sensor B and the color recognition result of the sensor A are the same.

The output of the color information signal selection unit 40 is given to the color information processing unit 70, and the brightness selection signal bsel is generated by the mode switching signal mode from the output switching unit 80. The brightness selection signal bsel is supplied to the brightness signal selection unit 9
Given to 0. As a result, the luminance signal selection unit 90
Based on the color information signal cl1 that has been color-determined by the color determination unit (for example, the color determination unit 31) of the class corresponding to the peak value peak determined by the level determination unit 60, for example, a reddish color (R
The signal rd) is selected and output to the brightness signal correction unit 100 as the brightness signal bw. The brightness signal correction unit 100 uses the peak value peak provided from the peak detection unit 50 as a basis.
The input luminance signal bw is corrected and output to the dropout processing unit 110. The dropout processing unit 110
Based on the attribute switching signal atr and the color determination signal col output from the color information processing unit 70, it is determined whether or not the pixel is a dropout color, the dropout color is removed, and the result is shown as image data img. Two
Output to the image memory 4. As described above, the color processing unit according to the first embodiment has four color determination units 31 to 34 classified into classes according to different signal level determination criteria.
And a level determination unit 6 that determines the output class of the color image sensor based on the output level of the color image sensor.
0. Then, the color information signal selection unit 40 selects the color information signals cl1 to cl4 of the color determination units 31 to 34 based on the class (color information selection signal csel) determined by the level determination unit 60, and the luminance selection signal bsel. Is getting Further, since the brightness signal is selected based on the brightness selection signal bsel, color recognition is performed uniformly without being affected by the difference in output level caused by the difference in individual characteristics (individual difference) of the color image sensor. It has the advantage that it can be done.

Second Embodiment FIG. 9 is a block diagram showing the configuration of a color processing unit in an OCR showing a second embodiment of the present invention, and common reference numerals are given to elements common to those in FIG. Is attached. In the color processing unit of this embodiment, a delay circuit 36 is added to the output side of the color discrimination units 31 to 34 of the color processing unit of FIG. 1, and the same function as the peak detection unit 50 is provided on the output side of the luminance signal selection unit 90. The peak detecting unit 91 having the same is added. Further, instead of the output of the peak detection unit 50 of FIG. 1, the brightness correction signal bcor output from the added peak detection unit 91 is applied to the brightness signal correction unit 100. With such a configuration, the peak detection unit 50 and the level determination unit 60
The delay time corresponding to the time for processing in the delay circuit 3
6, the color information selection unit 40 eliminates the time difference between the color information signals cl1 to cl4 and the color information selection signal csel. Also, the brightness signal correction unit 100
Also in the above, the brightness signal bw and the brightness correction signal bcor are input to the pixel selected by the brightness signal selection unit 90 at the same timing. Therefore, there is an advantage that the color determination and the luminance signal correction can be accurately performed for each pixel without being affected by the delay due to the processing time.

Third Embodiment FIG. 10 is a block diagram showing the arrangement of a color processing unit in an OCR showing a third embodiment of the present invention, and elements common to those in FIG. Is attached. In the color processing unit of this embodiment, a peak switching unit 50A is provided instead of the peak detection unit 50 of the color processing unit of FIG. The peak switching unit 50A includes a peak detection unit 50-
1, 50-2, 50-3 and peak maximum value detection unit 56
And a brightness level selection unit 57. The peak detectors 50-1, 50-2, and 50-3 have R, G, and B, respectively.
The signals rd, gr, bl are input respectively. The peak detectors 50-1 to 50-3 are the peak detectors 5 of FIG.
The output side of each of the peak detection sections 50-1 to 50-3 is connected to the peak maximum value output section 56 and the brightness level selection section 57, respectively. The peak maximum value output unit 56 includes the peak detection unit 50.
Peak values pk1 and pk input from -1 to 50-3
The maximum peak value of 2 and pk3 is the maximum peak value mxpk
Is output. The output side of the peak maximum value output unit 56 is connected to the brightness level selection unit 57.

The brightness level selecting section 57 includes a peak detecting section 5
Each peak value p input from 0-1 to 50-3
From k1 to pk3 and the maximum peak value mxpk input from the peak maximum value output unit 56, one is selected based on the mode selection signal mode given from the output switching unit 80 and output as the brightness level value blvl. To do. The output side of the brightness level selection unit 57 is the level determination unit 6
0. Since the peak switching unit 50A is configured in this way, the peak switching unit 50A performs the following operation. When the form 1 is read by the color image reading unit 2 of FIG. 2, the peak reset signal r is displayed over a plurality of lines immediately after the start of reading, as shown in FIG.
st is turned on. Thereby, the peak detector 50-1
50-3 are R, G, B signals rd, gr, b, respectively.
l peak is detected, and R signal peak value pk1,
The G signal peak value pk2 and the B signal peak value pk3 are held. The peak values pk1 to pk3 held in the peak detection units 50-1 to 50-3 are the peak maximum value output unit 56.
Given to. The peak maximum value output unit 56 selects the maximum value from the peak values pk1 to pk3 and outputs it as the maximum peak value mxpk. This maximum peak value mxp
k and the respective peak values pk1 to pk3 held in the peak detection units 50-1 to 50-3 are input to the brightness level selection unit 57. The brightness level selection unit 57 selects one of the input peak values pk1 to pk3 based on the mode selection signal mode given from the output switching unit 80.

The brightness level selecting section 57 selects the R signal peak value pk1 when the processing mode is "a mode excluding only red colors", and G when it is "a mode excluding only blue colors". Signal peak value pk2 or B signal peak value pk3
Select Further, in the "mode excluding all colors other than black", the maximum peak value mxpk is set to the brightness level value bl.
Select as vl. In this way, the brightness level value blvl selected by the brightness level selection unit 57 is output to the level determination unit 60. The operation other than the peak switching unit 50A is the same as that of the color processing unit of FIG. FIG.
It is explanatory drawing for demonstrating the effect of the color processing part of FIG. In FIG. 11, the peak reset signal rst is used to scan the blank area at the leading edge of the form to scan the peak values pk1 to pk1.
Each signal rd of R, G, B when pk3 is detected,
An example of the signal levels of gr and bl is shown. When scanning white, the peak value pk1 (= 22) of the R signal rd
It is assumed that the peak value pk2 (= 236) of the G signal gr is higher than 0). In this case, the maximum peak value mxp
When k (= 236) is used as the brightness level value blvl, the peak value pk of the G signal gr is calculated for pixels of bluish color.
Since the class determination is performed by the level determination unit 60 based on 2, no problem occurs. However, if the pixel is a reddish color pixel, the actual peak value pk1 of the R signal rd is class 2,
The level determination unit 60 has determined that the class is 1,
It becomes impossible to select the luminance signal in the regular class (that is, class 2).

In the color processing section of FIG. 10, the peak value p of the R, G, B signals rd, gr, bl in each of the peak detecting sections 50-1 to 50-3 in the peak switching section 50A.
k1 to pk3 are retained. Then, the mode selection signal m
When the ode is, for example, "mode excluding only red color", the peak value pk1 of the R signal rd is selected and given to the level determination unit 60, so the level determination unit 60 determines to be class 2. As a result, the luminance signal is selected in the regular class, and correct color recognition can be performed.
On the contrary, when the peak value pk1 of the R signal rd is high and the "mode excluding only blue color" is specified, the peak value pk2, pk3 of the G signal gr or the B signal bl causes the level determination unit Since the class judgment is done,
Correct color recognition is performed for blue colors. Thus, in the third embodiment, the R, G, B signals rd, gr,
Peak detectors 50-1 to 50-3 are provided for each bl, and the maximum value of each peak value pk1 to pk3 is set to the maximum peak value mx.
A peak maximum value output unit 56 that outputs pk and a brightness level selection unit 57 that switches between the maximum peak value mxpk and each peak value pk1 to pk3 are provided. This allows
Even if there is a difference in the signal levels of the R, G, and B signals rd, gr, and bl, the luminance level can be selected with the regular peak value peak in the processing mode in which the single primary color pixel signal is used. . Therefore, there is an advantage that accurate image data img can be obtained.

Fourth Embodiment In the fourth embodiment of the present invention, the line type color image sensor in the color image reading section 2 of the OCR of FIG. 2 has a pixel array as shown in FIG. 4 (2). If so, the resolution is increased and the processing is performed. In the pixel arrangement of FIG. 4 (2), the red and blue color sensors are alternately arranged on the RB (red and blue) line, so that the number of red and blue pixels is only half that of the G (green) line. Absent. Therefore, in order to obtain a resolution equivalent to that of the G line, it is necessary to interpolate the primary color component signals of the pixel positions that do not exist from the primary color component signals of the same color on both sides. FIG. 12 is a block diagram showing a configuration of a color processing unit in the OCR showing the fourth embodiment of the present invention, and elements common to those in FIG. 10 are designated by common reference numerals.

The color processing unit of the fourth embodiment outputs the RB signal and the G signal output from the color image reading unit 2 of the line type color image sensor having a pixel arrangement as shown in FIG. 4B, for example. It has a sensor interface section 21A for inputting. Sensor interface unit 2
The output side of 1A is connected to the RGB alignment section 22A and the interpolation signal generation section 23. The RGB alignment unit 22A has R,
The G, B signals rd, gr, bl are sampled, and even-numbered pixels and odd-numbered pixels on the pixel array line of the line-type color image sensor are separately output. In addition, the interpolation signal generation unit 23 uses the RB signal rb supplied from the sensor interface unit 21A.
From the R signal rd and the B signal bl to interpolate the R signal r
i and the interpolation B signal bi. The output sides of the RGB alignment unit 22A and the interpolation signal generation unit 23 are connected to the even pixel processing unit 200 and the odd pixel processing unit 300 having the same configuration.

The even-numbered pixel processing section 200 has an even-numbered pixel among primary color component signals corresponding to the pixels on the form read by the line-type color image sensor line by line and the interpolation signal generated by the interpolation signal generation section 23. The signal corresponding to the pixel is processed. The odd-numbered pixel processing unit 300 processes a signal corresponding to an odd-numbered pixel. The even-numbered pixel processing unit 200 includes the color determination units 31 to 3 of the color processing unit of FIG.
4, maximum value detector 35, delay circuit 36, color information selector 4
0, peak detection units 50-1 to 50-3, peak maximum value output unit 56, brightness level selection unit 57, level determination unit 60,
The color information processing unit 70 and the luminance signal selecting unit 90 have substantially the same functions and configurations. That is, the even-numbered pixel processing unit 200 includes the color determination units 31e to 34e, the maximum value detection unit 35e, the delay circuit 36e, the color information selection unit 40e, the peak detection units 50-1e to 50-3e, and the peak maximum value output unit 56e. Brightness level selection unit 57e, level determination unit 60
e, a color information processing unit 70e, and a luminance signal selection unit 90e. The output side of the luminance signal selection unit 90e is connected to the data switching unit 91.

The odd pixel processing unit 300 and the even pixel processing unit 2 are also included.
00, the color determination units 31o to 3 each having the same function.
4o, maximum value detection unit 35o, delay circuit 36o, color information selection unit 40o, peak detection units 50-1o to 50-3o, peak maximum value output unit 56o, brightness level selection unit 57o, level determination unit 60o, color information processing The unit 70o and the luminance signal selection unit 90o are included. The output side of the brightness signal selection unit 90o is connected to the data switching unit 91. The data switching unit 91 alternately switches and outputs the brightness signal bw input from the brightness signal selection unit 90e and the brightness signal selection unit 90o. Data switching unit 91
The output side of the peak detection unit 95 and the luminance signal correction unit 10
0. The output side of the peak detection unit 95 is connected to the brightness signal correction unit 100, and the brightness signal correction unit 100 is connected.
The output side of is connected to the dropout processing unit 110.

FIG. 13 is a block diagram showing the structure of the RGB alignment section 22A shown in FIG. This RGB alignment section 22A
Is the RB input from the sensor interface unit 21A.
It has a register 22a for reading the signal rb in synchronization with the transfer clock clk1. The output side of the register 22a is connected to the register 22b, and the output side of the register 22b is connected to the register 22c. Furthermore, the register 22
The output sides of b and the register 22c are connected to the register 22d. These registers 22b, 22c, 22d
Is to read and output signals on the input side in synchronization with a transfer clock clk2 having a phase opposite to that of the transfer clock clk1. The R3 signal rd3 and the R1 signal rd1 which are odd-numbered R signals are output to the output side of the register 22d. The RB signal rb is given to the register 22e, and the output side of the register 22e is connected to the register 22f. Further, the output sides of the registers 22e and 22f are connected to the register 22g. These registers 22e, 22f, 22g are
The signal on the input side is read and output in synchronization with the transfer clock clk2. The B4 signal bl4 and the B2 signal bl2, which are even-numbered B signals, are output to the output side of the register 22g.

On the other hand, the G signal gr input from the sensor interface section 21A is read by the register 22h in synchronization with the transfer clock clk1. The output side of the register 22h is connected to the register 22i, and the output side of the register 22i is connected to the register 22j. Further, the G signal gr is input to the register 22k. The output side of the register 22k is connected to the register 22l, and the output side of the register 22l is connected to the register 22m.
These registers 22i, 22j, 22k, 22l, 2
2m is for reading and outputting the signal on the input side in synchronization with the transfer clock clk2. As a result, the output side of the register 22j outputs the G3 signal gr3 which is an odd-numbered G signal, and the output side of the register 22m outputs the G2 signal gr2 which is an even-numbered G signal.

FIG. 14 is a block diagram showing the structure of the interpolation signal generator 23 of FIG. This interpolation signal generation unit 23
Has a register 23a to which the RB signal rb is input from the sensor interface unit 21A. Register 23
The output side of a is connected to the register 23b, and the register 23b
The output side of b is connected to the register 23c. The registers 23a, 23b, and 23c respectively transfer clocks c
The signal on the input side is read and output in synchronization with a transfer clock clk3 having a frequency twice that of lk1. The output sides of the registers 23a and 23c are connected to the average data creation unit 23d. Average data creation unit 23
d is for calculating and outputting the average value of two input signals. The output side of the average data creation unit 23d is connected to the registers 23e and 23f. Register 2
The output side of 3e is connected to the register 23g, and the register 2
The output side of 3f is connected to the register 23h. The register 23f reads and outputs the signal on the input side in synchronization with the transfer clock clk1. Register 23
e, 23g, and 23h are for reading and outputting the signal on the input side in synchronization with the transfer clock clk2. On the output side of the register 23g, the interpolated even-numbered R signal r
i is output, and the interpolated odd-numbered B signal bi is output to the output side of the register 23h.

FIG. 15 shows the RGB alignment section 22 shown in FIG.
It is a figure for demonstrating operation | movement of A and the interpolation signal production | generation part 23 shown in FIG. The operation of the color processing unit shown in FIG. 12 will be described below with reference to FIGS. 13 to 15. FIG.
RB from the OCR color image reading unit 2 to the RGB alignment unit 22A through the sensor interface unit 21A.
Signal rb, G signal gr, and transfer clocks clk1, c
lk2 is input. In the RGB alignment section 22A of FIG. 13, the RB signal rb and the G signal gr are transferred to the transfer clock clk.
The data is read into the registers 22a and 22h in synchronism with 1 and read into the registers 22e and 22k in synchronism with the transfer clock clk2. Register 22b,
22c and 22d time-align the RB signal rb read in the register 22a in synchronization with the transfer clock clk2 and output it as odd-numbered R signals (R3 signal rd3 and R1 signal rd2). Similarly, register 22e
The read RB signal rb is output as an even-numbered B signal (B4 signal bl4 and B2 signal bl2). In addition, the register 22h and the register 22k receive the read G signal g
r is output as an odd-numbered G signal (G3 signal gr3) and an even-numbered G signal (G2 signal gr2), respectively.
15, RB signals read in synchronization with transfer clocks clk1, clk2 at times t1, t2, ...
The time relationship between the G signal and the R3, R1, B4, B2, G3, and G2 signals aligned and output by the RGB alignment unit 22A is shown.

On the other hand, in the interpolation signal generator 23 of FIG.
The RB signal rb is input and read into the register 23a in synchronization with the transfer clock clk3 having a frequency twice that of the transfer clock clk1. The registers 23b and 23c temporally align the RB signal rb read into the register 23a by the transfer clock ckl3, and output the RB signal rb to the average data creating unit 23d. The average data creation unit 23d calculates and outputs the average value of the two input signals as follows. Interpolated R signal = (R1 + R3) / 2 Interpolated B signal = (B2 + B4) / 2 Registers 23e, 23f, 23g, and 23h transfer clock clk1 so that these outputs have the same time as the output of RGB alignment unit 22A. , Clk2, and outputs as an interpolating R signal ri and an interpolating B signal bi. FIG. 15 shows the time relationship between the interpolated R signal ri and the interpolated B signal bi that are aligned and output by the interpolated signal generation unit 23. For example, at time t7 in FIG.
On the output side of the GB alignment unit 22A and the interpolation signal generation unit 23,
As an even-numbered pixel signal, (R1 + R3) / 2, G
The signals ri2, gr2 and bl2 of 2 and B2 are output. In addition, as odd-numbered pixel signals, R3, (B2 +
B4) / 2 and G3 signals rd3, bi3 and gr3 are output.

These even-numbered pixel signals are supplied to the even-numbered pixel processing section 200 of the color processing section of FIG. Further, the odd-numbered pixel signals are given to the odd-numbered pixel processing unit 300. Even pixel processing unit 200 and odd pixel processing unit 30
The difference of 0 is that the even pixel processing unit 200 uses the corrected R signal ri in place of the R signal rd, and the odd pixel processing unit 300 uses the corrected B signal bi in place of the B signal bl. Luminance signal selection unit 90 of the even pixel processing unit 200
The luminance signal bw output from e and the odd pixel processing unit 30
The brightness signal bw output from the brightness signal selection unit 90o of 0
Are given to the data switching unit 91. Data switching unit 91
Alternately switches between the two input luminance signals bw,
The signal is output to the peak detection unit 95 and the brightness signal correction unit 100. The brightness signal correction unit 100 corrects the input brightness signal bw based on the peak value peak given from the peak detection unit 95 and outputs the corrected brightness signal bw to the dropout processing unit 110. The dropout processing unit 110 includes a color information processing unit 70.
It is determined whether or not the pixel is a dropout color based on the attribute switching signal atr and the color determination signal col that are output from the dropout color, and the dropout color is removed. The result is used as the image data img in the image memory 4 of FIG. Output to.

As described above, in the fourth embodiment, red,
An interpolation signal generation unit 23 is provided in OCR color processing using a line-type color image sensor having two rows of green and blue primary color pixel arrays. Then, the signal at the pixel position where the R signal rd and the B signal bl do not exist is interpolated from the R signal rd and the B signal bl on both sides thereof to generate an interpolated R signal ri and an interpolated B signal bi. These interpolated signals ri, b
i is treated in the same manner as the normal R, G, B signals rd, gr, bl, and the peak value pk1 is calculated by the peak detectors 50-1 to 50-3.
~ Pk3 is detected, and the level determination unit 60 determines the input signal level. Then, based on the determination result, the color information selection unit 40 performs color determination, and the brightness signal selection unit 90 selects the brightness signal bw. Therefore, correct color determination is possible even when using a single-color pixel signal. Furthermore,
Since the interpolation signals ri and bi are also used as the image data img, the resolution can be improved and the character pattern can be reproduced more faithfully.

The present invention is not limited to the above embodiment, and various modifications can be made. For example, there are the following modifications. (A) Each of the color processing units in FIGS. 1, 9, 10, and 12 uses four color determination units 31 to 34, but a different number of color determination units may be used. . When the level of the color component signal input from the color image sensor is subdivided, more accurate determination can be performed by increasing the number of color determination units. (B) In the color processing unit of FIG. 12, the red and blue color signals are interpolated, but other colors may be interpolated according to the pixel arrangement of the color image sensor. (C) In the color processing unit of FIG. 12, even-numbered signals and odd-numbered signals are separately processed by the even-numbered pixel processing unit 200 and the odd-numbered pixel processing unit 300, and then combined by the data switching unit 91. However, it is also possible to perform processing by one processing unit. In this case, although a circuit for switching the input signal is required depending on whether each signal is an even number or an odd number, it is possible to avoid the duplication of the processing units and simplify the processing.

[0038]

As described in detail above, according to the first aspect of the invention, a plurality of color determining units that use different reference values and a level determining unit that determines the class of the input signal level from the peak value of the input signal level. Section is provided. This allows
The input color component signal and luminance signal can be selected according to the determination result of the color determination unit that uses the reference value corresponding to the class determined by the level determination unit. Therefore, without being affected by the output level of the color image sensor,
This has the effect of enabling stable character recognition. Second
In the invention, since the second peak detection unit for detecting the peak of the selected luminance signal is added to the configuration of the first invention, the correction signal for correcting the luminance signal and the temporal correction of the luminance signal to be corrected are performed. The gap disappears. As a result, in addition to the effect of the first invention, there is an effect that accurate color determination and accurate luminance signal correction can be performed in pixel units. In the third invention, in place of the first peak detector in the second invention,
Since the first peak detector that detects the peak values of the red, green, and blue component signals is provided, it is possible to select and use the peak value according to the processing mode. As a result, it is possible to perform more accurate color determination than the second aspect of the invention, and it is possible to perform more accurate character recognition. In the fourth invention, since the complementary signal generating unit for interpolating the primary color component signal is added to the configuration of the third invention, it is possible to increase the pixel density in a pseudo manner. As a result, in addition to the same effect as the third aspect of the invention, there is an effect that the resolution is improved and accurate character recognition is possible even when a color image sensor having a low pixel density is used.

[Brief description of drawings]

FIG. 1 is a block diagram showing a configuration of a color processing unit of an OCR showing a first embodiment of the present invention.

FIG. 2 is a block diagram showing a schematic configuration of a conventional OCR.

3 is a block diagram showing a configuration of a color processing unit of the conventional OCR of FIG.

FIG. 4 is a diagram showing an example of a pixel array pattern of a line type color image sensor.

FIG. 5 is a diagram for explaining a problem of the conventional color determination method.

FIG. 6 is a block diagram showing a configuration of a peak detection unit in FIG.

FIG. 7 is a diagram showing an example of a relationship between a reading position of a form and a peak reset signal.

8 is an explanatory diagram illustrating an example of processing of a level determination unit in FIG.

FIG. 9 is a block diagram showing a configuration of a color processing unit of the OCR showing the second embodiment of the present invention.

FIG. 10 is a block diagram showing a configuration of a color processing unit of the OCR showing the third embodiment of the present invention.

FIG. 11 is a diagram for explaining the effect of the color processing unit in FIG.

FIG. 12 is a block diagram showing a configuration of a color processing unit of an OCR showing a fourth embodiment of the present invention.

13 is a block diagram showing a configuration of an RGB aligning unit in FIG.

14 is a block diagram showing a configuration of an interpolation signal generation unit in FIG.

15 is an operation explanatory diagram of the RGB alignment unit of FIG. 13 and the interpolation signal generation unit of FIG.

[Explanation of symbols]

 1 Form 23 Interpolation signal generation part 31 to 34 Color determination part 35 Maximum value detection part 40 Color information selection part 50, 50-1 to 50-3, 95 Peak detection part 56 Peak maximum value output part 57 Brightness level selection part 60 Level Judgment unit 90 Luminance signal selection unit 100 Luminance signal correction unit bcor Luminance correction signal bi Interpolation B signal bl B signal blvl Luminance level value bsel Luminance selection signal bw Luminance signal cl1 to cl4 Color information signal col Color signal csel Color information selection signal gr G Signal img image data max maximum value signal mode mode selection signal mxpk maximum peak value rd R signal ri interpolated R signal peak, pk1 to pk3 peak value

Front Page Continuation (72) Inventor Koji Senoo 1-7-12 Toranomon, Minato-ku, Tokyo Oki Electric Industry Co., Ltd.

Claims (5)

[Claims] 1. A color image reading means for decomposing characters written on a form into pixels, optically reading them, and separating and outputting to primary color component signals of red, green, and blue, and between the respective primary color component signals. Color processing means for performing color determination based on the signal level difference and outputting the determined color component signal of red, green, blue or achromatic color as image data, and registration from the image data registered in the character recognition dictionary. In an optical character reader comprising a recognition means for recognizing a character written on the form with reference to a pattern, the color processing means uses different reference values for the signal level difference. A plurality of color determination units that perform color determination and output the determination results as color information signals, respectively, and a maximum output that detects and outputs the maximum value signal with the maximum signal level from the primary color component signals. A value detection unit, a color information selection unit that outputs a color information signal selected by a color information selection signal from the color information signals output from the plurality of color determination units, and a signal level of the maximum value signal A peak detection unit that detects the peak of the signal and holds and outputs it as a peak value, and determines the class of the signal level difference based on the peak value and outputs the color information selection signal according to the determination result. Section, a color information processing section for outputting a luminance selection signal based on the color signal and a mode selection signal given according to the processing mode, and each of the primary color component signals and the maximum value signal based on the luminance selection signal. A luminance signal selecting section for selecting one from the above and outputting it as a luminance signal; and a luminance signal for correcting the signal level of the luminance signal based on the peak value and outputting image data to the recognizing means. Tadashibu Prefecture the optical character reading apparatus characterized by comprising. 2. A color image reading means for separating characters written on a form into pixels and optically reading them and outputting them as red, green, and blue primary color component signals, and a signal level between the primary color component signals. A color processing unit that performs color determination based on the difference and outputs the determined red, green, blue, or achromatic color component signal as image data; and from the image data, refer to a registered pattern registered in a character recognition dictionary. Then, in the optical character reading device comprising a recognition unit for recognizing the characters written on the form, the color processing unit performs color determination using different reference values for the signal level difference. A plurality of color determination units that perform determination and output the determination results as color information signals, respectively, and a maximum value that detects and outputs the maximum value signal having the maximum signal level from the primary color component signals. An output unit, a color information selection unit that outputs a color information signal selected by a color information selection signal from among the color information signals output from the plurality of color determination units, and a signal level of the maximum value signal. A first peak detection unit that detects a peak, holds and outputs it as a peak value, and determines the class of the signal level difference based on the peak value, and outputs the color information selection signal according to the determination result. A level determination unit, a color information processing unit that outputs a luminance selection signal based on the color signal and a mode selection signal given according to the processing mode, and each of the primary color component signals and the maximum based on the luminance selection signal. A brightness signal selection unit that selects one of the value signals and outputs it as a brightness signal, and a second peak detection unit that detects the peak of the signal level of the brightness signal and holds and outputs it as a brightness correction signal. The luminance and a luminance signal correction section that outputs the image data to the recognizing means corrects the signal level of the luminance signal correction signal based on the optical character reader, characterized in that it includes. 3. A color image reading means for separating characters written on a form into pixels, optically reading them, and outputting them as red, green, and blue primary color component signals, and a signal level between the respective primary color component signals. A color processing unit that performs color determination based on the difference and outputs the determined red, green, blue, or achromatic color component signal as image data; and from the image data, refer to a registered pattern registered in a character recognition dictionary. Then, in the optical character reading device comprising a recognition unit for recognizing the characters written on the form, the color processing unit performs color determination using different reference values for the signal level difference. A plurality of color determination units that perform determination and output the determination results as color information signals, respectively, and a maximum value that detects and outputs the maximum value signal having the maximum signal level from the primary color component signals. An output unit, a color information selection unit that outputs a color information signal selected by a color information selection signal from among the color information signals output from the plurality of color determination units, and each of the red, green, and blue A first peak detection unit that detects a peak of a signal level for each primary color component signal, holds and outputs each as a peak value, and a maximum peak value among a plurality of peak values output from the first peak detection unit A peak-maximum value output section that outputs a thing as a maximum peak value; and a plurality of peak values output from the first peak detection section and the maximum peak value, depending on a mode selection signal given according to a processing mode. A brightness level selection unit that selects one and outputs it as a brightness level value, and a level that determines the class of the signal level difference based on the brightness level value and outputs the color information selection signal according to the determination result. A color determination section, a color information processing section that outputs a luminance selection signal based on the color signal and the mode selection signal, and one of the primary color component signals and the maximum value signal based on the luminance selection signal. A luminance signal selecting section for selecting one of the luminance signals and outputting it as a luminance signal; a second peak detecting section for detecting the peak of the signal level of the luminance signal and holding and outputting it as a luminance correction signal; An optical character reading device, comprising: a luminance signal correction unit that corrects the signal level of the luminance signal based on the output and outputs image data to the recognition unit. 4. A color image reading means for separating characters written on a form into pixels and optically reading them and outputting them as primary color component signals of red, green and blue, and a signal level between the respective primary color component signals. A color processing unit that performs color determination based on the difference and outputs the determined red, green, blue, or achromatic color component signal as image data; and from the image data, refer to a registered pattern registered in a character recognition dictionary. Then, in the optical character reading device comprising a recognition means for recognizing the characters described on the form, the color processing means, the primary color component signal of the pixel position where the primary color component signal does not exist, the pixel Interpolation signal generators that interpolate from the signal levels of the primary color component signals of the same color on both sides of the position to generate interpolated primary color component signals, and different standards for the signal level difference. A plurality of color determination units that perform color determination using the above, and output the determination results as color information signals, and a maximum value detection that detects and outputs a maximum value signal with a maximum signal level from each of the primary color component signals. Section, a color information selection section that outputs, as a color signal, one selected by a color information selection signal from among the color information signals output from the plurality of color determination sections, each of the primary colors of red, green, and blue. A first peak detection unit that detects a peak of a signal level for each of the component signal and the interpolation primary color component signal, holds and outputs each as a peak value, and a plurality of peak values output from the first peak detection unit Of the maximum peak value output unit that outputs the maximum value as the maximum peak value from among the plurality of peak values and the maximum peak value output from the first peak detection unit, depending on the processing mode. To be A brightness level selection unit that selects one by a mode selection signal and outputs it as a brightness level value; a class of the signal level difference is determined based on the brightness level value; and the color information selection signal is determined according to the determination result. A level determination unit for outputting, a color information processing unit for outputting a luminance selection signal based on the color signal and the mode selection signal, and based on the luminance selection signal, from among the respective primary color component signals or the respective primary colors A luminance signal selecting section that selects one of the component signal and the interpolation primary color component signal and outputs it as a luminance signal; a peak of the signal level of the luminance signal that is detected and held and output as a luminance correction signal; And a brightness signal correction unit for correcting the signal level of the brightness signal based on the brightness correction signal and outputting image data to the recognition unit. Optical character reader according to claim. 5. The number of classes of the signal level difference determined by the level determination unit is the same as the number of the plurality of color determination units, and the level determination unit uses each of the plurality of color determination units for color determination. 5. The optical character reader according to claim 1, wherein the class is determined according to the reference value of.