JP2000235620A - Character recognizing device - Google Patents

Abstract

PROBLEM TO BE SOLVED: To provide a character recognizing device which obtains a high quality grayscale image for recognition from one input image inputted by using a single plate color image sensor without depending upon a character or background colors. SOLUTION: This device is provided with an image inputting means 1 which obtains an input image by a single plate color image pickup device D constructed by arranging plural pixels in a multiple matrix, an image generating means 2 for character detection which defines a pixel corresponding to a specific single color in an input image as a reference pixel and prepares a grayscale image for character detection on the basis of the pixel value of the reference pixel, a character detecting means 3 which detects a character area from the grayscale image for character detection, an image producing means 4 for character recognition which produces a grayscale image for recognition by using a pixel value in a character area about the character area in the input image and a character recognizing means 5 which recognizes a character from the grayscale image for recognition.

Description

DETAILED DESCRIPTION OF THE INVENTION

[0001]

BACKGROUND OF THE INVENTION 1. Field of the Invention The present invention relates to a character recognition device for recognizing characters in an image, and more particularly to a character recognition device for recognizing characters in an input image using a single-plate color image sensor. is there.

[0002]

2. Description of the Related Art FIG.
FIG. 1 is a schematic configuration diagram showing a conventional character recognition device described in Japanese Unexamined Patent Publication (Kokai) No. H10-15095. In the figure, reference numeral 101 denotes an image input unit for inputting a color input image, 3 denotes a character detection unit for detecting characters from the obtained image, and 5 denotes a character recognition unit for recognizing characters from the obtained character area.

As detailed components of the image input means 101, A is an optical lens, B is an optical axis moving means, C is an optical axis moving control means, D is a single color image sensor, and E is an A / D.
Conversion means, F is first storage means, H is second storage means,
I is color / monochrome designating means, J is luminance component detecting means, K
Is a monochrome image generating means.

Next, the operation will be described. In such a conventional apparatus, the user sets whether the imaging target is a color image or a black-and-white image, and the operation differs depending on the instruction of the color / black-and-white specifying means I based on the user setting. A case of obtaining a black and white (gray scale) image on the premise of recognition will be described.

[0005] First, an optical image (reflected light) from a subject incident through an optical lens A is moved by an optical axis moving means B under the control of an optical axis movement control means C to form a single-plate color image. The A / D converter E converts the image signal generated by the single-color image sensor D into a digital image by applying the image signal to the sensor D, and stores the digital image in the first storage unit F. Further, since the instruction by the color / monochrome designating means I is a black and white image, in addition to the image taken without shifting the pixels, the image whose pixel position is shifted by half a pixel to the left, right and down by the optical axis moving means B To obtain a total of four screens. Note that the single-plate color image sensor D used in this example is the same as that shown in FIG.
As shown in FIG. 4, a complementary color pixel structure of a set of four pixels having different sensitivities for each wavelength band (color) of light is provided.

Next, the luminance component detecting means J detects the color / color
Based on the black and white image instruction by the black and white specifying means I,
For the screen, a luminance component is obtained independently from each color component of the filter. The luminance component is obtained by obtaining a normalization function in advance for each pixel corresponding to each color, and applying this normalization function to the imaged pixel value. The luminance components for the four screens obtained by this operation are converted into black and white image generating means K.
Are stored in a memory on the second storage means H corresponding to the shift position to form one grayscale image.

Thereafter, the character detecting means 3 detects individual characters in the grayscale image, and the character recognizing means 5 recognizes the detected character pattern. Regarding character detection and character recognition, conventional methods such as Mitsubishi Electric Technical Report
Vol. 62 No. 2 pp. The method described in 9-12 “License plate recognition technology” is used.

[0008]

In a conventional character recognition device using an image pickup device using a single-chip color image sensor D as image input means, the luminance component of each pixel is obtained by a predetermined normalization function. If the object is black and white (for example, the character is white and the background is black), or if the color of the character and background is known in advance, it is possible to prepare a normalization function corresponding to that color However, it is not possible to cope with the case where the color of the character or the background is unknown, and a high-quality grayscale image cannot be obtained. Therefore, there is a problem that recognition accuracy is reduced. In addition, since high resolution by pixel shift combines images captured a plurality of times, it is effective for a stationary target, but cannot be applied when the target is moving.

SUMMARY OF THE INVENTION The present invention has been made to solve the above-described problem. A gray scale image for character detection is created using only predetermined pixel values, and each detected character area is detected. By re-creating the grayscale image for recognition using the pixel values in the image, a high-quality grayscale image for recognition can be obtained from a single input image input using a single-chip color image sensor, regardless of the color of characters and background. It is an object of the present invention to provide a character recognition device for obtaining a scale image.

[0010]

The character recognition means according to the present invention comprises the following means. (A) image input means for obtaining an input image with a single-chip color imaging device in which a plurality of pixels are arranged in a matrix
(B) a character detection image creating means for creating a gray scale image for character detection based on a pixel value of the reference pixel, using a pixel corresponding to a specific single color in the input image as a reference pixel;
(C) character detection means for detecting a character area from a gray scale image for character detection, and (d) character recognition for generating a gray scale image for recognition using a pixel value in the character area for the character area in the input image. Image creation means; (e) character recognition means for recognizing characters from a grayscale image for recognition.

The character detecting image generating means generates a character detecting gray scale image using only the pixel value at the reference pixel position of the input image.

Further, the character detecting image creating means sets a value calculated from a pixel value of a peripheral reference pixel adjacent to each non-reference pixel in the input image as a pixel value of the non-reference pixel. A gray scale image for character detection is created based on the pixel values.

The character recognizing image creating means calculates, for each character area, a gray-scale correction ratio of the non-reference pixel, and calculates a pixel value of the reference pixel and a pixel value of the non-reference pixel corrected by the gray-scale correction ratio. A grayscale image for recognition is created based on.

The character recognizing image creating means calculates, for each character area, a gray-scale correction ratio of the non-reference pixel and a reference value corrected by the gray-scale correction ratio. A plurality of pixel values are calculated by different calculations from the pixel values of the pixels, and a value closest to the reference value is determined as the pixel value of the non-reference pixel from the plurality of pixel values, based on the pixel value and the pixel value of the reference pixel. To create a grayscale image for recognition.

The plurality of pixel values calculated by the calculation are the average value of the pixel values of the left and right reference pixels of the predetermined non-reference pixel and the average value of the pixel values of the upper and lower reference pixels.

The character recognizing image creating means obtains an average pixel value for each color for each character area, and sets a ratio of the non-reference pixel to the reference pixel in the average pixel value as a gray level correction ratio.

The image generating means for character recognition divides all the pixels into a plurality of groups according to the color and the magnitude of the pixel value for each character area, obtains an average pixel value for each group, and calculates an average pixel value of the group. The ratio of the non-reference pixel to the reference pixel is defined as the density correction ratio for each group.

Further, the single-chip color image sensor is a color image sensor of a primary color system, and the reference pixel is a pixel corresponding to green.

[0019]

DESCRIPTION OF THE PREFERRED EMBODIMENTS Embodiment 1 FIG. 1 is a schematic configuration diagram showing a character recognition device of the present invention. In FIG. 1, the same components as those in the related art are denoted by the same reference numerals, and description thereof is omitted.
1 is an image input means for inputting an image using a single-chip color image sensor D as a single-chip color image sensor, 2 is a character detection image creating means for creating a gray scale image for character detection, and 4 is a detected character. This is a character recognition image creating unit that creates a recognition grayscale image for a character region in the input image corresponding to the position.

As detailed components of the image input means 1,
A is an optical lens, D is the single color image sensor described above, E is A / D conversion means, and F is storage means.

Next, the operation will be described with reference to the drawings. FIG. 2 is a processing flowchart of the present invention. Here, the operation will be described with reference to FIGS. FIG. 3 shows a pixel array of a single-chip color image sensor, FIG. 4 shows an input image, FIG. 5 shows a gray scale image for character detection, FIG. 6 shows a detected character area in the gray scale image for character detection, and FIG. FIG. 4 is a diagram illustrating a detected character area in an image.

First, an image input means 1 using a single-chip color image sensor D inputs and digitizes an image to create an input image (step S1). The single-chip color image sensor is configured by arranging a plurality of pixels having sensitivity to a specific color (wavelength band of light). For example, the Journal of the Institute of Image Electronics Engineers of Japan Vol. 27, No. 2, (1998) p
p. As described in Table 3 of “2. Imaging device for digital still camera” of 83 to 89, there are two types of primary color system and complementary color system. In this embodiment, an example using a primary color system will be described. The primary color system has a configuration in which the four pixels shown in FIG. In FIG. 3, 6 and 9 are pixels having sensitivity to the G (green) component, and 7 is R
A pixel having sensitivity to the (red) component and a pixel 8 having sensitivity to the B (blue) component.

FIG. 4 shows an input image created by the image input means 1. In FIG. 4, the input image has a width of w pixels and a height of h pixels, and each pixel is represented by a corresponding color component and a coordinate value. For example, G (2, 4) represents a G component pixel at a position where the X coordinate is 2 and the Y coordinate is 4.

Next, the character detecting image forming means 2 forms a character detecting gray scale image from the input image of FIG. 4 (step S2). In this step, pixels of a specific color component in the input image are used as reference pixels. In the present embodiment, an example will be described in which a G component (green) pixel is used as a reference pixel. First, two adjacent vertical lines in the input image are set as a line pair, and only the G component pixel value is extracted from this line pair to form one vertical line. The pixels of this vertical line are combined to form a gray scale image for character detection.

FIG. 5 shows a gray scale image for character detection created from the input image of FIG. In FIG. 5, each pixel indicates a corresponding pixel in the input image. The image size is p pixels in width and h pixels in height, and the image width p is half the width w of the input image. As for the correspondence relationship with the input image, the vertical lines of X coordinates 1 and 2 in the input image of FIG. 4 correspond to the vertical lines of X coordinate 1 in FIG. 5, and the vertical lines of X coordinates 3 and 4 in the input image correspond to those in FIG. This corresponds to the vertical line at X coordinate 2.

Next, the character detecting means 3 detects characters from the gray scale image for character detection in the same procedure as in the prior art (step S3).

Thereafter, the character recognition image creating means 4 creates a recognition gray scale image for each detected character area (step S4). First, the coordinates of the detection position in the gray scale image for character detection are converted into coordinates in the input image, and a character area in the input image is obtained. FIG. 6 is a diagram showing an example of a detected character area in a gray scale image for character detection, and 10 is the character area. FIG. 7 shows a character area in the input image corresponding to the detected character area 10 in FIG. 6, and 11 is a character area in the input image. The gray scale image for character detection in the present embodiment is an image having the same height and half the width as the input image, so that the Y coordinate of the detection position is the same, but the X coordinate is different.

The character recognizing image creating means 4 calculates an average pixel value for each color component for the character area in the input image (step S41). By this operation, an average pixel value of each of the G component (green), the R component (red), and the B component (blue) is obtained. Next, the character recognizing image creating means 4 calculates the ratio of the average pixel value of the color component of the non-reference pixel (other than the reference pixel) to the average pixel value of the reference pixel to obtain a correction ratio (step S42). In the present embodiment, since the reference pixel is a G component pixel, the correction ratio of the R component and the B component is obtained according to the equation (1).

[0029]

(Equation 1)

Further, the character recognition image creating means 4 corrects the pixel value for the character area in the input image (step S4).
3) to create a grayscale image for recognition. In this pixel value correction, the value of the reference pixel is used without correction, and for the non-reference pixel, a gray scale value is obtained according to Equation 2 using a correction ratio corresponding to the corresponding color component.

[0031]

(Equation 2)

Next, the character recognizing means 5 recognizes a character from the gray scale image for recognition in the same procedure as in the prior art (step S5). The operations in steps S4 to S5 are repeated by the number of detected characters, and the process ends when the recognition of all the detected characters is completed (step S5).
6).

In this embodiment, by performing the above-described steps, a high-quality grayscale image can be recognized even when the colors of the characters and the background are unknown, and high recognition accuracy can be obtained. Further, since only one captured image is used, it is possible to cope with a moving object. Furthermore, creating a grayscale image for character detection for the entire input image is a simple method with a small amount of processing, and creating a grayscale image for recognition, which requires a large amount of processing for high quality, uses only the detected character area. By targeting
It is possible to realize a character recognition device that has high recognition accuracy while suppressing the processing amount and is excellent in cost performance.

Embodiment 2 Hereinafter, a second embodiment of the present invention will be described with reference to the drawings. FIG. 1 is a schematic configuration diagram similar to the first embodiment. Next, the operation will be described with reference to the drawings. FIG. 8 is a processing flowchart. Here, the operation will be described with reference to FIGS. FIG. 9 is a gray scale image for character detection.

First, in the same procedure as in the first embodiment, the image input means 1 inputs an image to create an input image (step T1). Next, the character detection image creation means 2 creates a character detection grayscale image from the input image (step T2). In the present embodiment, the reference pixel is the same G component pixel as in the first embodiment, but the gray scale image for character detection has the same width and height as the input image, and the value of the reference pixel is used as it is. The value of the reference pixel is determined from the surrounding reference pixel values to obtain a gray scale image for character detection.

FIG. 9 is an example of a gray scale image for character detection created corresponding to the input image of FIG. 4. In FIG. 9, G corresponds to a reference pixel in the input image, and this pixel value is the input image. Is the same as the pixel value. On the other hand, G ′ indicates a non-reference pixel in the input image, and this pixel value is calculated from peripheral reference pixel values. For example, a method of calculating the average value of the reference pixel values adjacent to the left and right, a method of calculating the average value of the reference pixel values of the upper and lower sides, and a method of calculating the average value of the reference pixel values of the upper and lower sides are used. However, in the present embodiment, a method will be described in which the average value of the reference pixel values adjacent to the left and right is used. In this case, for example, G ′ in FIG.
The pixel value of (3,2) is an average value of the pixel values of G (2,2) and G (4,2).

Thereafter, the same procedure as in the first embodiment is used.
Character detection is performed (step T3), and a grayscale image for recognition and character recognition are performed for all detected character regions (steps T4 to T6).

In this embodiment, through the above steps, a high-resolution gray scale image for character detection can be created, and the character detection accuracy can be improved. In this embodiment, the pixel value of the gray scale image for character detection corresponding to the non-reference pixel is the average value of the reference pixel values adjacent to the left and right. Other methods such as a method of setting a value may be used.

Embodiment 3 Hereinafter, a third embodiment of the present invention will be described with reference to the drawings. FIG. 1 is a schematic configuration diagram similar to the first embodiment. Next, the operation will be described with reference to the drawings. FIG. 10 is a processing flowchart. Here, the operation will be described with reference to FIG. FIG. 11 is a diagram for explaining pixel value correction.

First, in the same procedure as in the first embodiment, the image input means 1 inputs an image to create an input image (step U1), and the character detecting image creating means 2 uses the character detecting grayscale from the input image. A scale image is created (step U2), and the character detection means 3 detects a character from the gray scale image for character detection (step U3).

Next, the character recognition image creating means 4 creates a recognition gray scale image for the detected character area in the input image (step U4). In this operation, as in the first embodiment, first, the average pixel value is calculated for each color component (step U41), and the correction ratio of the non-reference pixel is calculated for each color component (step U42). Further, the character recognition image creating means 4 corrects the value of the non-reference pixel with the correction ratio, but in the present embodiment, the corrected value is not directly used as the pixel value of the recognition grayscale image.
A reference value for pixel value complement is set (step U43).

The pixel value complement of the non-reference pixel using the reference value (step U44) will be described below. In this complementation, the complementation value of the pixel to be complemented is determined from the value of the peripheral reference pixel, and the value closest to the reference value is selected from among a plurality of complementation values calculated by a plurality of different methods to complement the pixel. Value, and this is used as the pixel value of the grayscale image for recognition. There are various methods for the plurality of complementing methods. In the present embodiment, there are two methods: a method of averaging the values of vertically adjacent reference pixels and a method of averaging the values of left and right adjacent reference pixels. The case where one is used will be described. FIG. 11 is a diagram for explaining this pixel value complementation. Reference numeral 12 denotes a non-reference pixel to be complemented, reference numerals 13 to 16 denote reference pixels adjacent to the non-reference pixel 12, and reference pixels 13 to 16 in FIG. Numerical values indicate the value of each pixel in the input image. Here, first, the character-recognition-image creating unit 4 determines the average value (120 + 1) of the values of the reference pixels vertically adjacent to each other as the first complement value.
50) ÷ 2 = 135, and calculate the average value of the values of the reference pixels adjacent to the left and right (52 + 68) ÷ 2 = 60 as the second complement value. Next, the absolute value of the difference between the reference value of the non-reference pixel 12 obtained in step U43 and the two complementary values is calculated, and the complementary value that minimizes this value, that is, the complementary value closest to the reference value, is selected. . For example, the reference value of the non-reference pixel 12 is 1
Assuming that the absolute value of the difference from the first complement value is ABS
(120-135) = 15, the absolute value of the difference from the second complement value is ABS (120-60) = 60, and the first complement value is selected. In the above description, ABS indicates an operator that takes an absolute value.

Thereafter, character recognition is performed in the same procedure as in the first embodiment (step U5). Further, the above-described generation of the grayscale image for recognition and the character recognition are repeated until there is no longer any character area that has not been subjected to the recognition processing (step U
4-U6).

In this embodiment, through the above-described steps, a high-quality gray scale image for recognition can be created even when the average pixel value of a specific color component is small and the correction by the correction ratio has a large error. In the present embodiment, as a method of complementing non-reference pixels in creating a grace case image for character recognition, a method of taking an average value of values of reference pixels adjacent vertically and a method of taking an average value of reference pixels adjacent to left and right are used. Although two methods have been described, this may be another method, or three or more methods may be applied.

Embodiment 4 Hereinafter, a fourth embodiment of the present invention will be described with reference to the drawings. FIG. 1 is a schematic configuration diagram similar to the first embodiment. Next, the operation will be described with reference to the drawings. FIG. 12 is a processing flowchart.

First, in the same procedure as in the first embodiment, the image input means 1 inputs an image to create an input image (step V1), and the character detection image creating means 2 uses the character detection grayscale from the input image. A scale image is created (Step V2), and the character detecting means 3 detects a character from the gray scale image for character detection (Step V3).

Next, the character recognition image creating means 4 creates a recognition gray scale image for the character area in the input image (step V4). In this step, first, pixels are grouped for each color component according to the pixel value (step V41). In this grouping, for example, a binarization level is obtained for all pixels of a specific color component, and a pixel having a value larger than this binarization level belongs to the color component group 1 and binarization is performed. Pixels having a value equal to or lower than the level belong to the color component group 2 and are divided into two groups. By performing this operation for each of the R, G, and B components, a total of six groups can be obtained. In determining the binarization level, for example, the IEICE Transactions V
ol. J63-D No. 4 pp 349-356, “Automatic Threshold Selection Method Based on Discrimination and Least Square Criterion”.

Next, the image generating means 4 for character recognition calculates an average pixel value for each group (step V42), and further calculates a correction ratio for each group (step V43).
In calculating the correction ratio, the correction ratio of each group of the non-reference pixels is obtained according to, for example, Equation 3.

[0049]

(Equation 3)

Further, the character recognition image creating means 4 calculates a reference value for the non-reference pixel by multiplying the correction ratio of the group to which the pixel belongs by the value of the pixel (step V44). Thereafter, in the same manner as in the third embodiment, a complement value is obtained from the peripheral reference pixel value and the reference value for the non-reference pixel (step V4).
5) After creating the gray scale image for recognition, the character recognition means 5 performs character recognition (step V5).

In this embodiment, through the above steps, a high-quality grayscale image for recognition can be created even when the ratio of the color components is extremely different between the character portion and the background portion. In this embodiment, the pixels are grouped into two groups for each color component, but may be three or more groups. Further, although the criterion for grouping is the binarization level, other methods may be used.

In the first to fourth embodiments, the color image sensor is of a primary color system and the reference pixel is a G component pixel. However, a single color image sensor other than the primary color system may be used. The reference pixel may be a pixel other than the G component. In addition, as an operation flow, the process from the generation of the grayscale image for recognition to the character recognition is repeated by the number of the detected character regions, but this is performed by creating the grayscale image for recognition for all the character regions and then performing character recognition collectively. You may do it.

[0053]

The character recognition means according to the present invention includes the following means. (A) image input means for obtaining an input image with a single-chip color imaging device in which a plurality of pixels are arranged in a matrix
(B) a character detection image creating means for creating a gray scale image for character detection based on a pixel value of the reference pixel, using a pixel corresponding to a specific single color in the input image as a reference pixel;
(C) character detection means for detecting a character area from a gray scale image for character detection, and (d) character recognition for generating a gray scale image for recognition using a pixel value in the character area for the character area in the input image. Image creation means; (e) character recognition means for recognizing characters from a grayscale image for recognition. Therefore, the grayscale image for character detection is created using the pixel values of the predetermined pixels.However, by recreating the grayscale image for recognition using the pixel values in the detected individual character regions, the character A high-quality grayscale image for recognition can be obtained from an image input by a single-chip color image sensor regardless of the background color or the color of the background, and high recognition accuracy can be realized.

The character detecting image forming means forms a character detecting gray scale image using only the pixel values at the reference pixel positions of the input image. Therefore, by extracting a reference pixel from the input image to create a grayscale image for character detection, a grayscale image for character detection can be easily created, and high-speed character detection can be performed as the image size is reduced. .

The character recognizing image creating means calculates, for each character area, a gray-scale correction ratio of the non-reference pixel and a reference value corrected by the gray-scale correction ratio. A plurality of pixel values are calculated by different calculations from the pixel values of the pixels, and a value closest to the reference value is determined as the pixel value of the non-reference pixel from the plurality of pixel values, based on the pixel value and the pixel value of the reference pixel. To create a grayscale image for recognition. Therefore, the value of the reference pixel is used as it is,
By determining the value of the non-reference pixel from the value of the reference pixel around the pixel and using it as a character detection grayscale image,
Character detection can be performed on an image that maintains the resolution of the input image.

Further, the character recognizing image creating means calculates, for each character area, the gray-scale correction ratio of the non-reference pixel, and calculates the pixel value of the reference pixel and the pixel value of the non-reference pixel corrected by the gray-scale correction ratio. A grayscale image for recognition is created based on. Therefore, by correcting the value of the non-reference pixel in each character region to create a recognition grayscale image, it is possible to obtain a high quality recognition grayscale image in which the pixel value levels of the character portion and the background portion are uniform. it can.

The character recognizing image generating means calculates, for each character area, a gray-scale correction ratio of the non-reference pixel and a reference value corrected by the gray-scale correction ratio, and further calculates a different value from the pixel value of the peripheral reference pixel. The pixel value of the non-reference pixel adjacent to the reference value is calculated as the pixel value of the adjacent non-reference pixel from the plurality of pixel values, and the gray value for recognition is determined based on the pixel value and the pixel value of the reference pixel. Create a scale image. Therefore, for a non-reference pixel, the grayscale correction value is used as a reference value, and the recognition grayscale value is determined from the pixel values of the reference pixels around the pixel. A high-quality grayscale image for recognition with a uniform pixel value level in the background can be obtained.

The plurality of pixel values calculated by the calculation are the average value of the pixel values of the left and right reference pixels of the predetermined non-reference pixel and the average value of the pixel values of the upper and lower reference pixels.
Therefore, for a non-reference pixel, the grayscale correction value is used as a reference value, and a value closest to the reference value is selected from among the average value of the left and right reference pixels and the average value of the upper and lower reference pixels, and is used for recognition of the pixel. By using the gray scale value, correction using the reference pixel value can be realized by a simple neighborhood operation.

Further, the character recognizing image generating means obtains an average pixel value for each color for each character area, and sets the ratio of the non-reference pixel to the reference pixel in the average pixel value as the gray level correction ratio.
Therefore, in each character area, the average pixel value for each corresponding color is determined, and the ratio of the average pixel value is set as the density correction ratio of the pixel corresponding to each color, so that the pixel value level of the character portion and the background portion is uniform. A high-quality grayscale image for recognition can be obtained.

Further, the character recognizing image creating means divides all the pixels into a plurality of groups according to the color and the magnitude of the pixel value for each character area, obtains an average pixel value for each group, and obtains the average pixel value. The ratio of the non-reference pixel to the reference pixel is defined as the density correction ratio for each group. Therefore, in each character area, all the pixels are divided into a plurality of groups according to the corresponding color and the pixel value, and an average pixel value is obtained for each group, and the ratio of the average pixel value is determined by the density correction ratio of the pixels belonging to each group. By doing so, a high-quality grayscale image for recognition can be obtained even when the ratio of the color components is extremely different between the character portion and the background portion.

Further, the single-chip color imaging device is a color image sensor of a primary color system, and the reference pixel is a pixel corresponding to green. Therefore, the image input means inputs an image using a primary color color image sensor, and the character detection image creation means creates a character detection grayscale image using a pixel corresponding to green in the input image as a reference pixel. Accordingly, a high-quality grayscale image for recognition can be obtained even when a primary-color-based color image sensor having good sensitivity for color signals but poor sensitivity for luminance signals is used.

[Brief description of the drawings]

FIG. 1 is a schematic configuration diagram showing a character recognition device of the present invention.

FIG. 2 is a processing flowchart of the character recognition device of the present invention.

FIG. 3 is a diagram showing a pixel arrangement of a single-chip color image sensor.

FIG. 4 is a diagram showing an input image of a single-chip color image sensor.

FIG. 5 is a diagram showing a gray scale image for character detection.

FIG. 6 is a diagram illustrating a detected character area in a gray scale image for character detection.

FIG. 7 is a diagram illustrating a detected character area in an input image.

FIG. 8 is a processing flowchart showing another example of the character recognition device of the present invention.

FIG. 9 is a diagram showing a gray scale image for character detection.

FIG. 10 is a processing flowchart showing another example of the character recognition device of the present invention.

FIG. 11 is a diagram for explaining pixel value correction.

FIG. 12 is a processing flowchart showing another example of the character recognition device of the present invention.

FIG. 13 is a schematic configuration diagram showing a conventional character recognition device.

FIG. 14 is a diagram illustrating a complementary-color pixel structure of a set of four pixels.

[Explanation of symbols]

1 image input means, 2 character detection image creation means, 3
Character detecting means, 4 character recognizing image creating means, 5 character recognizing means, D single-chip color image sensor (single-chip color image sensor).

 ──────────────────────────────────────────────────続 き Continued on the front page F term (reference) 5B029 AA01 BB02 CC21 DD09 5C065 AA07 BB30 CC01 CC04 DD01 EE12 GG26 GG32 5C066 AA01 BA01 CA27 DC06 EC12 GA04 GB01 HA02 HA03 JA01 KD01 KD06

Claims (9)

[Claims] 1. A character recognition device comprising the following means. (A) image input means for obtaining an input image with a single-chip color imaging device in which a plurality of pixels are arranged in a matrix, (b) pixels corresponding to a specific single color in the input image are set as reference pixels, (C) character detection means for detecting a character area from the character detection grayscale image based on the pixel value of the reference pixel; (d) the input; (E) character recognition means for recognizing a character from the recognition grayscale image for the character region in the image using a pixel value in the character region to generate a recognition grayscale image; . 2. The character detecting image generating unit according to claim 1, wherein the character detecting image generating unit generates the character detecting grayscale image using only the pixel value of the input image at the reference pixel position. Character recognition device. 3. The image detection means for character detection, comprising: a pixel value at a predetermined reference pixel position in the input image;
2. The character detection grayscale image is created based on a pixel value at a non-reference pixel position calculated from a pixel value of a reference pixel adjacent to the non-reference pixel.
Character recognition device according to the description. 4. The character recognition image creating means calculates, for each of the character areas, a gray-scale correction ratio of a non-reference pixel, and a pixel value of the reference pixel and a non-standard pixel corrected by the gray-scale correction ratio. The character recognition device according to claim 1, wherein a recognition grayscale image is created based on the pixel value of the character. 5. The image generating means for character recognition calculates, for each of the character regions, a gray-scale correction ratio of a non-reference pixel and a reference value corrected with the gray-scale correction ratio, and further calculates, for each non-reference pixel, A plurality of pixel values are calculated by different calculations from the pixel values of the surrounding reference pixels, and a value closest to the reference value is determined as the pixel value of the non-reference pixel from the plurality of pixel values. 4. A grayscale image for recognition is created based on the values.
The character recognition device according to any one of the above. 6. The plurality of pixel values calculated by the calculation are an average value of pixel values of left and right reference pixels of a predetermined non-reference pixel and an average value of pixel values of upper and lower reference pixels. The character recognition device according to claim 5, wherein 7. The image generating means for character recognition calculates an average pixel value for each color for each character area, and sets a ratio of the non-reference pixel to a reference pixel of the average pixel value as the density correction ratio. The character recognition device according to any one of claims 4 to 6, wherein: 8. The image generating means for character recognition divides all pixels into a plurality of groups for each character area according to the color and the magnitude of the pixel value, and calculates an average pixel value for each group. 7. The character recognition device according to claim 4, wherein a ratio of a non-reference pixel to a reference pixel of a value is set as the gray-scale correction ratio for each group. 9. The character according to claim 1, wherein the single-chip color imaging device is a primary color color image sensor, and the reference pixel is a pixel corresponding to green. Recognition device.