JP6191286B2 - Character recognition apparatus, character recognition method, and computer program for character recognition - Google Patents

Description

  The present invention relates to a character recognition device, a character recognition method, and a character recognition computer program for recognizing characters appearing on an image, for example.

  In order to recognize characters (including symbols and numbers) shown in the image, it is preferable that the character region in which the character is shown and the background region in which the background around the character is shown can be accurately identified. Therefore, for example, the character recognition device determines a binarization threshold based on the luminance value distribution of the image, and a pixel having a luminance value equal to or higher than the binarization threshold and a pixel having a luminance value less than the binarization threshold. By binarizing, the character area and the background area are identified. However, a shadow of another object may be cast on a subject on which characters to be recognized are written, for example, a license plate. In such a case, the image of the subject includes a character region, a background region, and a region where a shadow and a character or a shadow and a background are reflected. For this reason, even if the background is uniform, the distribution shape of luminance values is a multimodal shape. As a result, it becomes difficult to set an appropriate binarization threshold, and the character area and the background area may not be accurately separated. Therefore, the grayscale image is converted into a binary image using various binarization thresholds for the grayscale image, and the optimum binarization threshold is determined using the character recognition result, reliability, and position for each binarization threshold. Has been proposed (see, for example, Patent Document 1).

JP-A-8-153163

  However, depending on how the shadow is applied, the character recognition device may not be able to identify the character area and the background area with one binarization threshold. For example, in an image in which the brightness of the background area is higher than the brightness of the character area, if the brightness of the shadowed area is lower than the brightness of the character area, the character recognition device uses a single binarization threshold to Cannot be distinguished from other areas. If the character region is not correctly identified, the character recognition device may erroneously recognize the character recognition result by, for example, misrecognizing a shaded portion in the background as a part of the character.

  Therefore, an object of the present specification is to provide a character recognition device capable of accurately recognizing a character from an image obtained by photographing the region even when a region including the character is shaded.

  According to one embodiment, a character recognition device is provided. This character recognition device divides a histogram into a plurality of histogram peaks, and generates a histogram of luminance values of pixels in a predetermined area including characters on an image. A division unit including a maximum value frequency value, a plurality of histogram peaks, each of which is classified into one of a first group and a second group, and a histogram peak classified into the first group A combination of histogram peaks classified into two groups is changed, and for each combination, a pixel having a luminance value corresponding to a histogram peak classified into the first group among the pixels in the predetermined region is first. On the other hand, a pixel having a luminance value corresponding to the peak of the histogram classified in the second group is set as the second pixel value. A binary image generating unit that generates a plurality of value images, and for each of the plurality of binary images, detects a character reflected in the binary image, calculates the likelihood of the character, And a recognition unit that sets a character having the highest certainty among characters detected from each character as a character included in a predetermined area.

The objects and advantages of the invention will be realized and attained by means of the elements and combinations particularly pointed out in the appended claims.
It should be understood that both the foregoing general description and the following detailed description are exemplary and explanatory and are not restrictive of the invention as claimed.

  The character recognition device disclosed in the present specification can accurately recognize a character from an image obtained by capturing the area even if the area including the character is shaded.

(A) is a figure showing an example of a picture of a subject with a shadow. (B) is a figure which shows the histogram of the luminance value of the image shown by (a). (C) is a figure which shows an example of the binary image obtained by binarizing the image shown by (a) with one binarization threshold value. (D) converts the image shown in (a) into a set of pixels having luminance corresponding to the histogram peaks corresponding to the character area and a set of pixels having luminance corresponding to the other histogram peaks. It is a figure which shows an example of the binary image obtained by digitizing. It is a schematic block diagram of the character recognition apparatus by one Embodiment. It is a block diagram which shows the function of a process part. It is a figure which shows an example of the triangle which uses the set of a luminance value and frequency as one point. It is a figure which shows an example of the relationship between the binarization threshold value and the generated binary image when there are three peaks in the histogram. It is an operation | movement flowchart of a character recognition process. It is a figure which shows the correspondence of a character peak, a background area | region, the area | region where the shadow is applied to the character, and the area | region where the shadow is applied to the background, and the peak of a histogram in the case where there are four histogram value peaks.

Hereinafter, a character recognition device according to an embodiment will be described with reference to the drawings.
First, an example of an image in which a character area cannot be distinguished from other areas with only one binarization threshold will be described. FIG. 1A is a diagram illustrating an example of an image of a subject with a shadow. In the image 100, black characters are depicted on a white background. Further, the curved area 101 with shadows other than characters is also blacker than the surrounding background.

  FIG. 1B is a diagram showing a histogram of luminance values of the image shown in FIG. In FIG. 1B, the horizontal axis represents luminance, and the vertical axis represents frequency. The histogram peaks 111 to 114 respectively correspond to the luminance distribution of the shadowed area of the character, the luminance distribution of the shadowed area of the background, the luminance distribution of the character area, and the luminance distribution of the background area. Here, it is assumed that the binarization threshold Th is set between the peaks 113 and 114 of the histogram so as to distinguish the background area and other areas having the highest luminance value.

  FIG. 1C is a diagram illustrating an example of a binary image obtained by binarizing the image illustrated in FIG. 1A with the binarization threshold Th illustrated in FIG. is there. In this example, the binary image 120 is binarized into a non-shadowed background area and other areas, and it is difficult to distinguish between a shadowed area and a character area. In this example, since the brightness of the character area is higher than the brightness of the shadowed area, even if the binarization threshold Th is set between the peaks 112 and 113 of the histogram, The image is binarized into a shaded area and a non-shadowed area. Therefore, the binarization is not performed so that only the character area can be distinguished from other areas. In this example, an image is binarized so that pixels having luminance included in the histogram peak 111 or histogram peak 113 are distinguished from pixels having luminance included in the histogram peak 112 or histogram peak 114. Is preferable. By binarizing the image in this manner, the character area 131 is distinguished from the background area 132 in the binary image 130 as shown in FIG. Therefore, the possibility that a region with a shadow on the background is erroneously recognized as part of a character is reduced.

  Therefore, this character recognition apparatus obtains a histogram of luminance values of an image obtained by photographing a subject including characters, and divides the histogram for each peak of the histogram. This character recognition device classifies each peak of the histogram into either the first group or the second group while changing the combination. For each combination, the character recognition device obtains a value of a pixel having a luminance value corresponding to the peak of the histogram classified into the first group and a luminance value corresponding to the peak of the histogram classified into the second group. A plurality of binary images are generated by making the pixel values different from each other. The character recognition device executes character recognition for each binary image, and sets the character corresponding to the most probable recognition result as the character shown in the image.

  FIG. 2 is a schematic configuration diagram of a character recognition device according to an embodiment. The character recognition device 1 recognizes at least one character appearing in an image based on an image generated by the imaging unit 2 capturing a subject including at least one character to be recognized. For this purpose, the character recognition device 1 includes an image acquisition unit 11, an output unit 12, a storage unit 13, a storage medium access device 14, and a processing unit 15. Furthermore, the character recognition device 1 may have a display device such as a liquid crystal display for displaying the character recognition result. The processing unit 15 is connected to the image acquisition unit 11, the output unit 12, the storage unit 13, and the storage medium access device 14 via, for example, a bus.

The imaging unit 2 is a digital video camera, for example, and is installed so that the subject is included in the shooting range. For example, when the subject including characters to be recognized is a license plate of a vehicle, the imaging unit 2 is installed outdoors so that, for example, the front surface of the vehicle traveling on a road is included in the shooting range. Then, the imaging unit 2 shoots the subject to generate an image showing the subject. The imaging unit 2 outputs the image to the image acquisition unit 11 each time an image is generated.
In the present embodiment, the image generated by the imaging unit 2 is a gray image whose pixel value has only a luminance value.

The image acquisition unit 11 includes a communication interface for connecting to the imaging unit 2 and its control circuit. Such a communication interface can be, for example, an interface according to a communication standard for connecting peripheral devices such as Universal Serial Bus (USB), or an interface for video signal input.
Alternatively, the image acquisition unit 11 may include a communication interface for connecting to a communication network in accordance with a communication standard such as Ethernet (registered trademark) and its control circuit. In this case, the image acquisition unit 11 acquires an image from the imaging unit 2 via the communication network.
The image acquisition unit 11 stores the image in the storage unit 13 every time an image is acquired.

The output unit 12 includes, for example, a communication interface for connecting the subject recognition apparatus 1 to another device and its control circuit. Such a communication interface can be an interface according to a communication standard for connecting peripheral devices such as USB or a communication standard such as Ethernet (registered trademark).
The output unit 12 receives information representing the recognition result of the subject by the processing unit 15 from the processing unit 15 and outputs the information to another device. Note that the image acquisition unit 11 and the output unit 12 may be integrated.

  The storage unit 13 includes, for example, a readable / writable semiconductor memory and a read-only semiconductor memory. And the memory | storage part 13 memorize | stores the computer program run on the process part 15, the various information used in order to recognize the character reflected in the image, and the various data produced | generated in the middle of a character recognition process. . The storage unit 13 may store the image received from the image acquisition unit 11 until the character recognition process is completed.

  The storage medium access device 14 is a device that accesses the storage medium 16 such as a magnetic disk, a semiconductor memory card, and an optical storage medium. For example, the storage medium access device 14 reads a computer program for character recognition executed on the processing unit 15 stored in the storage medium 16 and passes the computer program to the processing unit 15 or stores it in the storage unit 13.

  The processing unit 15 includes one or a plurality of processors and their peripheral circuits. Then, the processing unit 15 recognizes characters that appear in the image received from the image acquisition unit 11.

  FIG. 3 is a block diagram illustrating functions of the processing unit 15. The processing unit 15 includes a cutout unit 21, a histogram generation unit 22, a division unit 23, a binary image generation unit 24, and a recognition unit 25. Each of these units included in the processing unit 15 is, for example, a functional module implemented by a computer program executed on a processor included in the processing unit 15. Alternatively, each of these units included in the processing unit 15 may be mounted on the character recognition device 1 as an integrated circuit in which circuits for realizing the respective functions are integrated.

  The cutout unit 21 cuts out a recognition target area including a character area and a surrounding background area for each character from the image. In this embodiment, the cutout unit 21 is disclosed in Y. Ariki and T. Teranishi, “Indexing and classification of TV news articles based on telop recognition”, Proc. Of 4th ICDAR, pp. 422-497, 1997. As described above, the recognition target area is cut out by paying attention to the feature that many edge components are detected from the recognition target area.

  The cutout unit 21 first detects an edge pixel whose luminance value spatially changes using an edge detection filter such as a prewitt filter or a sobel filter. Then, the cutout unit 21 counts the number of edge pixels for each pixel row in the horizontal direction and the vertical direction. For example, the cutout unit 21 obtains the total number of edge pixels for each block including a plurality of continuous pixel rows in the horizontal direction corresponding to the assumed character height on the image, and the total number is equal to or greater than a predetermined number. It is determined that a character string is included in the block. Then, the cutout unit 21 sets the upper and lower boundaries of the block including the character string as the upper end and the lower end of the recognition target area, respectively. Similarly, for each block including a character string, the cutout unit 21 calculates the total number of edge pixels for each sub-block including a plurality of continuous vertical pixel columns corresponding to the character width assumed on the image. Ask for. Then, the cutout unit 21 determines that one character is included in the sub-block whose total number is equal to or greater than a predetermined value. Then, the cutout unit 21 sets the left and right boundaries of the sub-block including the characters as the left end and the right end of the recognition target area. Thereby, the recognition object area | region containing the character is cut out for every character.

  Note that the cutout unit 21 may cut out the recognition target region from the image by using another method for cutting out a region including characters.

  The histogram generation unit 22 examines the luminance value of each pixel included in the recognition target region for each recognition target region, and obtains the number of pixels having the luminance value for each luminance value, that is, the frequency, thereby obtaining the luminance value. Generate a histogram of Note that the histogram generation unit 22 generates a pixel having a luminance value included in the band for each band having a predetermined luminance value width (for example, 2 to 8 when the luminance value is represented by 0 to 255). The number may be obtained as a frequency. Then, the histogram generation unit 22 stores the luminance value histogram in the storage unit 13.

  The dividing unit 23 divides the histogram into a plurality of histogram peaks. Each of the peaks in the histogram includes a maximum value of the frequency of luminance values.

  For this purpose, the dividing unit 23 calculates a skew value indicating a bias in the distribution of the histogram with the analysis range from the minimum value to the maximum value of the luminance value as disclosed in, for example, Japanese Patent Application Laid-Open No. 8-221513. . When the histogram is biased toward a lower luminance value, the dividing unit 23 resets the luminance value from the minimum value to the average value as the analysis range, and checks the skew value again. On the other hand, when the histogram is biased toward a higher luminance value, the dividing unit 23 resets the luminance value from the average value to the maximum value as the analysis range, and checks the skew value again. The dividing unit 23 binarizes the average value of the luminance values in the analysis range when the skew value becomes a value corresponding to a small distribution deviation of the histogram so as to distinguish the peak and the peak of the histogram. The threshold is used. The dividing unit 23 repeats the above processing using each of the analysis ranges divided into two by the binarization threshold as a new analysis range. When the sum of the frequencies between two adjacent binarization thresholds is less than a predetermined frequency threshold (for example, 1/10 of the total number of pixels in the recognition target area), the dividing unit 23 obtains the value immediately before that. Discard the binarization threshold value. Then, the dividing unit 23 sets a portion between two consecutive binarization threshold values obtained before that as one histogram peak.

  Alternatively, when the skew value within the analysis range becomes a value corresponding to a small bias in the distribution of the histogram, the dividing unit 23 maximizes the frequency for a luminance value lower than the average of the luminance values in the analysis range. A first luminance value is obtained. Similarly, the dividing unit 23 obtains a second luminance value having a maximum frequency for a luminance value higher than the average of luminance values in the analysis range. The dividing unit 23 obtains a third luminance value having a minimum frequency between the first luminance value and the second luminance value. Then, the dividing unit 23 sets the luminance value and the frequency corresponding to each of the first luminance value to the third luminance value as the coordinates of one point, with the luminance value as the horizontal axis and the frequency as the vertical axis. The area of a triangle connecting the coordinates of these three points is calculated.

  FIG. 4 is a diagram illustrating an example of a triangle having a set of luminance value and frequency as one point. In FIG. 4, the horizontal axis represents the luminance value, and the vertical axis represents the frequency. A line 400 represents a histogram, and points 401 to 403 are points representing the coordinates of a set of frequencies for the first to third luminance values and the luminance values, respectively. The dividing unit 23 calculates the area of the triangle 410 whose vertices are the points 401 to 403, respectively.

  If the area is larger than the predetermined threshold, the dividing unit 23 sets the third luminance value as the binarization threshold. The predetermined threshold is experimentally determined in advance, for example. When the area is larger than the predetermined threshold, the dividing unit 23 may determine the binarization threshold by applying, for example, the Otsu binarization method within the analysis range. On the other hand, if the area is less than the predetermined threshold, the dividing unit 23 determines that the entire analysis range is included in one histogram peak without using the third luminance value as the binarization threshold.

  The dividing unit 23 may obtain each histogram peak and the luminance value having the maximum frequency at the peak according to another method of distinguishing the histogram for each peak. Then, the dividing unit 23 applies the Otsu's binarization method for each pair of adjacent two histogram peaks by using the Otsu binarization method with the luminance range between the two histogram peaks as the analysis range. A binarization threshold value may be set between two histogram peaks.

  The dividing unit 23 stores, in the storage unit 13, a binarization threshold value that represents the boundary between the peaks of each histogram for each recognition target region.

  The binary image generation unit 24 classifies each of the plurality of histogram peaks into either the first group or the second group. At that time, the binary image generation unit 24 creates a plurality of combinations of histogram peaks by changing various combinations of histogram peaks classified into the first group and histogram peaks classified into the second group. . Then, for each combination of histogram peaks, the binary image generation unit 24 selects pixels having luminance values corresponding to the histogram peaks classified into the first group among the pixels in the recognition target region as the first pixels. Value (for example, 0). On the other hand, the binary image generation unit 24 sets a pixel having a luminance value corresponding to the peak of the histogram classified into the second group as the second pixel value (for example, 255). Thereby, the binary image generation unit 24 generates a plurality of binary images for each recognition target area. Note that the binary image generation unit 24 may create binary images for all possible combinations, or may create binary images only for some of all possible combinations.

  FIG. 5 is a diagram illustrating an example of a relationship between a binarization threshold value and a generated binary image when there are three histogram peaks. An example of a histogram of luminance values is shown on the upper side of FIG. In this histogram, the horizontal axis represents the luminance value, and the vertical axis represents the frequency. The peaks 501 to 503 of the histogram correspond to a background area with shadows, a character area, and a background area without shadows, respectively. The binarization thresholds Th1 and Th2 are set between the histogram peaks 501 and 502 and between the histogram peaks 502 and 503, respectively. On the other hand, an example of a generated binary image is shown on the lower side of FIG.

For example, the binary image 511 is generated by binarizing the recognition target region so that the value of a pixel having a luminance value lower than the binarization threshold Th1 is lower than the value of a pixel having another luminance value. Is done. Conversely, the binary image 512 is obtained by binarizing the recognition target area so that the value of a pixel having a luminance value lower than the binarization threshold Th1 is higher than the value of a pixel having another luminance value. Generated.
Also, the binary image 513 is recognized so that the value of a pixel having a luminance value equal to or higher than the binarization threshold Th1 and lower than the binarization threshold Th2 is lower than the value of a pixel having another luminance value. It is generated by binarizing the target area. Conversely, in the binary image 514, the value of a pixel having a luminance value that is equal to or higher than the binarization threshold Th1 and lower than the binarization threshold Th2 is higher than the value of a pixel having another luminance value. It is generated by binarizing the recognition target area.
Further, the binary image 515 is generated by binarizing the recognition target area so that the value of the pixel having the luminance value equal to or higher than the binarization threshold Th2 is lower than the value of the pixel having the other luminance value. Is done. Conversely, the binary image 516 binarizes the recognition target area so that the value of the pixel having the luminance value equal to or higher than the binarization threshold Th2 is higher than the value of the pixel having the other luminance value. Generated.

  The binary image generation unit 24 stores the generated binary images in the storage unit 13 for each recognition target area.

  The recognition unit 25 performs character recognition processing on each of the binary images for each recognition target area, detects characters appearing in each binary image, and obtains its certainty. And the recognition part 25 determines with the most probable character among the characters detected from each of the binary image as the character reflected in the recognition object area | region.

For each recognition target area, the recognition unit 25 performs template matching between each of the binary images obtained for the recognition target area and a plurality of templates corresponding to individual characters to be recognized. And the recognition part 25 calculates the normalization cross correlation value of a recognition object area | region and each template, for example. And the recognition part 25 calculates | requires the maximum value of a normalization cross-correlation value for every binary image of a recognition object area | region as probability that the character corresponding to the template used as the maximum value is reflected in the recognition object area | region, The character of the template corresponding to the maximum value is obtained.
The recognizing unit 25 compares the probabilities obtained for the respective binary images, and recognizes the character having the maximum certainty as a character appearing in the recognition target area.
The recognition unit 25 recognizes the character string shown in the image by arranging the characters recognized for each recognition target area according to the arrangement of the recognition target areas.

  Note that the recognizing unit 25 may detect the character shown in the binary image of the recognition target area and obtain the certainty according to another method for recognizing the character shown in the image. For example, the recognition unit 25 extracts a feature amount related to a character shape such as a character intersection and an end point from each binary image of the recognition target region. And the recognition part 25 inputs the feature-value extracted from the binary image into the discriminator which outputs the certainty that it is a predetermined character for every binary image, and it is the predetermined image to a binary image. You may ask for the certainty that the letter is reflected. And the recognition part 25 should just make the character with the largest probability as the detected character for every binary image, and let the maximum value of the certainty be the character certainty about the binary image. Such a discriminator can be, for example, a multilayer perceptron, a support vector machine, or a Real adaBoost discriminator, and is prepared in advance for each character to be recognized. And these discriminators are based on a so-called supervised learning algorithm corresponding to the discriminator based on a plurality of sample images in which characters to be recognized are captured in advance and a plurality of sample images in which the characters are not captured. To be learned.

  The processing unit 15 outputs information representing the character string recognized by the recognition unit 25 to another device via the output unit 12. At that time, the processing unit 15 may output information used for the recognition together with information representing the recognized character string to another device via the output unit 12. Alternatively, the processing unit 15 may store information representing the character string recognized by the recognition unit 25 in the storage unit 13.

  FIG. 6 is an operation flowchart of character recognition processing executed by the processing unit 15. The processing unit 15 executes this character recognition process for each image.

  The cutout unit 21 of the processing unit 15 cuts out a recognition target area for each character from the image (step S101). The histogram generation unit 22 of the processing unit 15 generates a histogram of luminance values for each recognition target area (step S102). The dividing unit 23 of the processing unit 15 divides the luminance value histogram for each peak of the histogram (step S103).

  The binary image generation unit 24 of the processing unit 15 classifies the peaks of each histogram into two groups for each recognition target region while changing the combination. Then, the binary image generation unit 24 generates a plurality of binary images by assigning different values for each group to pixels having luminance values corresponding to the peaks of the histogram belonging to the group (step S104). And the recognition part 25 of the process part 15 detects the character reflected in the binary image for every binary image about each authentication object area | region, and calculates | requires the certainty (step S105). The recognition unit 25 sets a character detected from the binary image having the maximum certainty among the respective binary images as a character reflected in the recognition target area (step S106). Then, the recognition unit 25 obtains a character string shown in the image by arranging characters recognized for each recognition target region in accordance with the arrangement order of each recognition target region, and outputs information representing the character string to the output unit 12. (Step S107). Then, the processing unit 15 ends the character recognition process.

  As described above, this character recognition device divides the histogram of the luminance value of the recognition target area on the image into a plurality of histogram peaks, and classifies each histogram peak into two groups while changing the combination. To do. This character recognition device generates a plurality of binary images by assigning different values for each group to pixels having luminance values corresponding to the peaks of the histogram belonging to the group. The character recognition apparatus performs character recognition for each binary image, and sets the character detected from the binary image that has obtained the most probable recognition result as the character actually reflected in the image. For this reason, this character recognition device appropriately distinguishes the character area and the background area even when the histogram of luminance values shows a multi-modal shape having three or more peaks due to a shadow on the subject area. Character recognition can be performed based on the segmented binary image. Therefore, this character recognition device can accurately recognize characters in the image.

  Next, a character recognition device according to a second embodiment will be described. The character recognition device according to the second embodiment suppresses the number of binary images to be generated in consideration of the characteristic of luminance change due to shadows when there are four histogram peaks.

  The character recognition device according to the second embodiment differs from the character recognition device according to the first embodiment in the processing of the binary image generation unit 24 of the processing unit 15. Therefore, the binary image generation unit 24 will be described below. For other components of the character recognition device according to the second embodiment, refer to the description of the corresponding components of the character recognition device according to the first embodiment.

If there are four peaks in the luminance value histogram for the recognition target area, if the first value or the second value is assigned to each pixel having a luminance value included in the peak, The value image generation unit 24 generates 2 ⁴ = 16 binary images at the maximum.
Here, when there are four histogram peaks, each histogram peak has a shadowed character region, a shadowed character region, a non-shadowed background region, and a shadowed region. It is estimated that it corresponds to any of the background areas. In addition, the luminance of the pixels included in the shaded area decreases. Therefore, the relationship between each histogram peak and the above four types of areas is limited to the four patterns shown in the table 700 of FIG.

In the table 700 shown in FIG. 7, the pattern 1 and the pattern 2 indicate the assumed relationship between the peak of the histogram and each region in the case of so-called white characters in which the luminance of the character region is higher than the luminance of the background region. Show. On the other hand, pattern 3 and pattern 4 show the assumed relationship between the peaks of the histogram and each region when the luminance of the character region is lower than the luminance of the background region. In addition, each of the ranks 1 to 4 represents a rank from the highest luminance value of the four histogram peaks. “Character + shadow region” represents a shadowed character region, and “background + shadow region” represents a shadowed background region.
Here, in the character template used by the recognition unit 25, pixels included in the character area have a first pixel value (for example, 0), and pixels around the character area have a second pixel value (for example, 255). ). In this case, in order to generate a binary image so that the normalized cross-correlation value of the template of the same character as the character shown in the binary image is maximized, the pixels included in the character region are the first even in the binary image. It is preferable that a pixel having a pixel value of 1 and a pixel included in the background area has a second pixel value.

  In the case of the pattern 1, the first pixel value is assigned to the pixel having the luminance value corresponding to the peak of the histogram of rank 1 or 2, and the second pixel is assigned to the pixel having the luminance value corresponding to the peak of the histogram of rank 3 or 4. Assign pixel values. In the case of pattern 2, the first pixel value is assigned to the pixel having the luminance value corresponding to the peak of the histogram of rank 1 or 3, and the pixel having the luminance value corresponding to the peak of the histogram of rank 2 or 4 is assigned to the first pixel value. Allocate 2 pixel values.

  Similarly, in the case of pattern 3, a second pixel value is assigned to a pixel having a luminance value corresponding to the peak of the histogram of rank 1 or 2, and a pixel having a luminance value corresponding to the peak of the histogram of rank 2 or 4 is assigned. A first pixel value is assigned. In the case of the pattern 4, the second pixel value is assigned to the pixel having the luminance value corresponding to the peak of the histogram of rank 1 or 3, and the pixel having the luminance value corresponding to the peak of the histogram of rank 2 or 4 is assigned to the second pixel value. A pixel value of 1 is assigned. Thereby, a binary image in which the character area and the background area are separated is obtained.

  As described above, when there are four histogram peaks, any one of the above-described patterns 1 to 4 is satisfied, and therefore the binary image generation unit 24 only needs to generate a maximum of four types of binary images. . In addition, when the recognition unit 25 performs character recognition based on the feature amount such as the intersection or end point position or number of characters, either the character region or the background region may have high brightness. Therefore, in this case, the binary image generation unit 24 may generate only two of the binary image corresponding to the pattern 1 or 3 and the binary image corresponding to the pattern 2 or 4.

As described above, the character recognition device according to the second embodiment takes into consideration the decrease in luminance in the shadowed area, and the histogram histogram peaks that may be reflected in the characters and the histogram peaks that are not in the shadowed regions. Only binary images corresponding to combinations classified into different groups are generated. Therefore, since this character recognition device can suppress the number of binary images to be generated, the amount of calculation of the entire character recognition process can be suppressed.
Even when the histogram of luminance values has five or more peaks, the binary image generation unit may distinguish the character area from the background area in consideration of the cause of such a luminance distribution. By generating only a certain binary image, the number of generated binary images can be suppressed.
Further, when the number of peaks in the histogram is three or less, the binary image generator 24 may generate binary images for all combinations that classify the peaks in the histogram into two groups. Then, when the number of peaks in the histogram is four or more, the binary image generation unit 24 may generate a binary image for only a part of the combination.

  Note that in each of the above-described embodiments or modifications, the image generated by the imaging unit 2 may be a color image represented in the RGB color system. In this case, the processing unit 15 converts the color image into, for example, an HLS color system image, and executes each process of the processing unit 15 based on the luminance value of each pixel of the converted image. Good.

  All examples and specific terms listed herein are intended for instructional purposes to help the reader understand the concepts contributed by the inventor to the present invention and the promotion of the technology. It should be construed that it is not limited to the construction of any example herein, such specific examples and conditions, with respect to showing the superiority and inferiority of the present invention. Although embodiments of the present invention have been described in detail, it should be understood that various changes, substitutions and modifications can be made thereto without departing from the spirit and scope of the present invention.

DESCRIPTION OF SYMBOLS 1 Character recognition apparatus 2 Imaging part 11 Image acquisition part 12 Output part 13 Storage part 14 Storage medium access apparatus 15 Processing part 16 Storage medium 21 Cutout part 22 Histogram generation part 23 Dividing part 24 Binary image generation part 25 Recognition part

Claims (5)

A histogram generation unit that generates a histogram of luminance values of pixels in a predetermined area including characters on the image;
The histogram is divided into a plurality of histogram peaks, each of the histogram peaks including a maximum value of luminance value frequency,
Each of the plurality of histogram peaks is classified into one of a first group and a second group, and a histogram peak classified into the first group and a histogram classified into the second group Changing the combination of peaks, for each of the combinations, out of the pixels in the predetermined area, the pixel having a luminance value corresponding to the peak of the histogram classified into the first group is set as the first pixel value, A binary image generation unit that generates a plurality of binary images of the predetermined region by setting a pixel having a luminance value corresponding to a peak of the histogram classified into the second group as a second pixel value;
For each of the plurality of binary images, a character appearing in the binary image is detected, and the likelihood of the character is obtained. Among the characters detected from each of the plurality of binary images, the certainty A recognition unit that sets the character having the highest likelihood as a character included in the predetermined area;
A character recognition device.   The binary image generation unit classifies, among the plurality of histogram peaks, a mountain in which a character may appear, into the first group, and classifies another mountain into the second group. The character recognition apparatus according to claim 1, wherein only a binary image corresponding to a combination of histogram peaks is generated.   When the number of peaks in the histogram is four, the binary image generation unit classifies the first histogram peak and the third histogram peak into a first group in order from the highest luminance value, And the binary image corresponding to the combination which classified the peak of the 2nd histogram and the peak of the 4th histogram into the 2nd group, the peak of the 1st histogram, and the peak of the 2nd histogram are the 1st group. The character recognition device according to claim 2, further generating a binary image corresponding to a combination in which the third histogram peak and the fourth histogram peak are classified into the second group. Generate a histogram of the luminance values of the pixels in a given area that contains characters on the image,
The histogram is divided into a plurality of histogram peaks, each of the histogram peaks including a local maximum of luminance frequency;
Each of the plurality of histogram peaks is classified into one of a first group and a second group, and a histogram peak classified into the first group and a histogram classified into the second group Changing the combination of peaks, for each of the combinations, out of the pixels in the predetermined area, the pixel having a luminance value corresponding to the peak of the histogram classified into the first group is set as the first pixel value, Generating a plurality of binary images of the predetermined region by setting a pixel having a luminance value corresponding to a peak of the histogram classified into the second group as a second pixel value;
For each of the plurality of binary images, a character appearing in the binary image is detected, and the likelihood of the character is obtained. Among the characters detected from each of the plurality of binary images, the certainty The character with the highest likelihood is the character included in the predetermined area,
Character recognition method. Generate a histogram of the luminance values of the pixels in a given area that contains characters on the image,
The histogram is divided into a plurality of histogram peaks, each of the histogram peaks including a local maximum of luminance frequency;
Each of the plurality of histogram peaks is classified into one of a first group and a second group, and a histogram peak classified into the first group and a histogram classified into the second group Changing the combination of peaks, for each of the combinations, out of the pixels in the predetermined area, the pixel having a luminance value corresponding to the peak of the histogram classified into the first group is set as the first pixel value, Generating a plurality of binary images of the predetermined region by setting a pixel having a luminance value corresponding to a peak of the histogram classified into the second group as a second pixel value;
For each of the plurality of binary images, a character appearing in the binary image is detected, and the likelihood of the character is obtained. Among the characters detected from each of the plurality of binary images, the certainty The character with the highest likelihood is the character included in the predetermined area,
A computer program that causes a computer to execute the operation.