JP2014164529A - Character recognition device, character recognition method, and program - Google Patents

Abstract

PROBLEM TO BE SOLVED: To achieve both the improvement in accuracy of character recognition and the reduction of computational complexity when performing character recognition processing on a grayscale image.SOLUTION: A character recognition device is provided which performs character recognition on a grayscale image including a character string comprising characters arrayed in a specific direction. The character recognition device first recognizes Ncharacters (Nis a prescribed value equal to or larger than 2) in order from an end of the character string. In character recognition for the Ncharacters, a plurality of character areas having various character area sizes are set in respective positions of the grayscale image, and character recognition results and evaluation values are calculated for respective character areas. Character recognition results of the Ncharacters and the most suitable character area size are determined on the basis of the calculated evaluation values. In character recognition of remaining characters, the character recognition device sets a character area having the most suitable character area size to respective positions of the gray scale image to perform character recognition processing.

Description

  The present invention relates to a character recognition device, a character recognition method, and a program, and more particularly, to a character recognition device, a character recognition method, and a program for performing character recognition on a grayscale image.

  A technique for performing character recognition on a picked-up image picked up by an image pickup apparatus has been applied to various uses including license plate recognition. In character recognition for a captured image, generally, processing for generating a binarized image from the captured image is performed, and character recognition processing is further performed for the binarized image. Such a technique is disclosed in, for example, Japanese Patent Application Laid-Open No. 2006-338578.

  One problem with the technique of generating a binarized image from a captured image and performing character recognition processing on the binarized image is that a good binarized image cannot be obtained if the image quality of the captured image is poor. is there. For example, in recent license plate recognition apparatuses, imaging with low light quantity and low resolution is performed in accordance with the cost reduction of the imaging apparatus. When the amount of light decreases, shadows occur, and the luminance distribution becomes non-uniform as the S / N ratio decreases. In addition, when the resolution is lowered, the character feature amount in the captured image is deteriorated. In such a situation, depending on the current binarization processing, a binarized image in which characters are broken is generated (for example, a binarized image in which character lines are lost) is generated, and the accuracy of character recognition decreases. End up. Such a problem can be avoided by using an imaging device that can capture a high-quality captured image, but an imaging device that can capture a high-quality captured image is generally expensive and disadvantageous in terms of cost reduction.

  In order to cope with the problem that a good binarized image cannot be obtained if the image quality of the captured image is poor, the inventors have performed a character recognition process without binarization, that is, a grayscale image ( We are investigating a technique for performing character recognition processing on grayscale images. For example, Patent Document 2 (Japanese Patent Laid-Open No. 2008-251029) and Non-Patent Document 1 (Kawahiro Sawa et al. “High accuracy of handwritten Kanji recognition using gradient feature vector and variable absorption covariance matrix”. ”, Journal of the Institute of Electronics, Information and Communication Engineers, D-II, Vol. J82-D-II, No. 11). Patent Document 2 discloses a technique for improving character recognition accuracy by obtaining a high-resolution image by a super-resolution processing method in character recognition processing. Non-Patent Document 1 discloses a technique for identifying a character class by statistical pattern recognition using a feature quantity such as a luminance gradient obtained from a grayscale image.

  In character recognition processing for a grayscale image, it is preferable to simultaneously improve the accuracy of character recognition and reduce the amount of calculation. However, according to the inventor's study, there is room for improvement in the above-described prior art.

JP 2006-338578 A JP 2008-251029 A

Kazuhiro Sawa et al. "High accuracy of handwritten Kanji recognition using gradient feature vector and variable absorption covariance matrix", IEICE Transactions D-II, Vol.J82-D-II, No.11

  Accordingly, an object of the present invention is to provide a technique for simultaneously improving the accuracy of character recognition and reducing the amount of calculation in character recognition processing for a grayscale image.

In one aspect of the present invention, a character recognition device is provided that performs character recognition on a grayscale image in which a character string composed of characters arranged in a specific direction is displayed. Character recognition apparatus, among the strings, N _C-number (N _C is 2 or more predetermined value) from the first end includes a first character recognition means for performing character recognition for the first character is a character, character Second character recognition means for performing character recognition on a second character, other than N _C characters, from the first end of the column. In the character recognition for the first character, the first character recognition means sets a plurality of first character regions having various character region sizes at respective positions in the gray scale image, and performs character recognition for each of the plurality of first character regions. Processing is performed to calculate a character recognition result and an evaluation value, and based on the calculated evaluation value, a character recognition result of the first character and a first optimum character area size that is an optimum character area size are determined. To do. The second character recognition means sets a plurality of second character areas having the first optimum character area size at each position of the gray scale image, performs character recognition processing for each of the plurality of second character areas, and performs a character recognition result. And the evaluation value, and the character recognition result of the second character is determined based on the character recognition result and the evaluation value calculated for each of the plurality of second character regions.

  At this time, the first character recognizing means determines each character position of the first character based on the evaluation value, determines an optimal character size that is an optimal character spacing from the determined character position, The two-character recognition means preferably sets the positions of the plurality of second character areas according to the optimum inter-character size.

In one embodiment, the first character recognition unit ranks the character recognition results obtained for each position of the plurality of first character regions based on the evaluation value. First character recognition means, for each character area size, ranking a character recognition results obtained for each position of the plurality of first character region has a higher evaluation value from among the character recognition result is one of m _C 1 (m _C is a predetermined number greater than N _C ) is extracted, and a first average value that is an average value of evaluation values of m _C character recognition results is calculated. It is preferable to determine the character region size having the highest average value as the optimum character region size.

In this case, for each character area size, the first character recognizing means includes pixels having luminance values corresponding to the evaluation values of m _C character recognition results, corresponding to the m _C character recognition results. An evaluation value-luminance value conversion plot image, which is an image located at a region position, is generated and a binarization process is performed on the evaluation value-luminance value conversion plot image to generate a binarized image. preferable. The first character recognition means performs a labeling process on the binarized image to give a label to each set of continuous pixels having the same color, and for each label, character recognition corresponding to a pixel belonging to the label A second average value that is an average value of the evaluation values of the results is calculated, and the first character recognizing unit sequentially calculates the centroid position of each label and the centroid position of other labels in order from the label having the second highest average value. The process of removing each label is repeated until N _C labels are left when the distance between the labels is calculated and the center of gravity has a center of gravity such that the distance from the center of gravity of another label is equal to or less than a predetermined threshold. Do. The first character recognizing means selects, for each of the remaining N _C labels, the pixel having the highest evaluation value of the corresponding character recognition result as the pixel belonging to each of the N labels, and has the highest evaluation value The pixel position is determined as the character position of the first character, and the optimal character size, which is the optimal character spacing, is determined from the determined character position. The second character recognition means sets the positions of the plurality of second character areas according to the optimum inter-character size.

In one embodiment, the character recognition device further includes third character recognition means for performing character recognition on the third character that is N _C characters from the second end opposite to the first end of the character string. N _C characters from the second end of the character string may include a fourth character recognition unit that performs character recognition on a fourth character that is a character other than the character, and a final character recognition result determination unit. . In the character recognition for the third character, the third character recognition means sets a plurality of third character areas having various character area sizes at respective positions in the gray scale image, and performs character recognition for each of the plurality of third character areas. Processing is performed to calculate a character recognition result and an evaluation value, and based on the calculated evaluation value, a character recognition result for the third character and a second optimum character area size that is an optimum character area size are determined. To do. The fourth character recognition means sets each position of the gray scale image of the plurality of fourth character areas having the second optimum character area size, performs character recognition processing for each of the plurality of fourth character areas, and performs a character recognition result. And the evaluation value, and the character recognition result of the fourth character is determined based on the calculated evaluation value. The final character recognition result determining means is a method in which the character recognition result of the specific character of the character string obtained by the first character recognition means and the second character recognition means is obtained by the third character recognition means and the fourth character recognition means. When the character recognition result matches the character recognition result, the matching character recognition result is determined as the final character recognition result of the specific character. If they do not match, the final character recognition result determination means is obtained by the character recognition result of the specific character obtained by the first character recognition means and the second character recognition means, and by the third character recognition means and the fourth character recognition means. The character recognition result having the higher evaluation value among the character recognition results of the specific character is determined as the final character recognition result of the specific character.

  In one embodiment, the first character recognition means includes storage means for storing similar character data in which similar characters are registered, and a character recognition result obtained by performing character recognition processing for each of the plurality of first character areas. Of these, the optimum character area size is determined without using the character recognition result matching the similar character and its evaluation value.

In another aspect of the present invention, there is provided a character recognition device that performs character recognition on a grayscale image on which a character string composed of characters arranged in a specific direction is displayed. The character recognition apparatus, in the character string, N _C-number (N _C is 2 or more predetermined value) from the edge includes a first character recognition means for performing character recognition for the first character is a character of the string Second character recognition means for performing character recognition on the second character, which is a character other than N _C characters, from the end of. In the character recognition for the first character, the first character recognition means sets a plurality of first character regions having various character region sizes at respective positions in the gray scale image, and performs character recognition for each of the plurality of first character regions. A character recognition result and an evaluation value are calculated by performing processing, and the character recognition result of the first character and each character position of the first character are determined based on the calculated evaluation value, and the determined character position To determine the optimum character spacing, which is the optimum character spacing. The second character recognition means sets a plurality of second character areas at each position of the grayscale image according to the optimum inter-character size, performs character recognition processing for each of the plurality of second character areas, The evaluation value is calculated, and the character recognition result of the second character is determined based on the character recognition result and the evaluation value calculated for each of the plurality of second character regions.

In still another aspect of the present invention, there is provided a character recognition method for performing character recognition on a grayscale image in which a character string composed of characters arranged in a specific direction is displayed. The character recognition method, in the character string, N _C-number (N _C is 2 or more predetermined value) from the edge includes the step of performing character recognition for the first character is a letter, N from the end of the string Performing character recognition for a second character other than _C characters. In character recognition for the first character, a plurality of first character regions having various character region sizes are set at respective positions in the gray scale image, and character recognition processing is performed for each of the plurality of first character regions. A result and an evaluation value are calculated, and based on the calculated evaluation value, a character recognition result of the first character and a first optimum character area size which is an optimum character area size are determined. In character recognition for the second character, a plurality of second character areas having a first optimum character area size are set at each position of the grayscale image, and character recognition processing is performed for each of the plurality of second character areas. The character recognition result and the evaluation value are calculated, and the character recognition result of the second character is determined based on the character recognition result and the evaluation value calculated for each of the plurality of second character areas.

In still another aspect of the present invention, there is provided a character recognition method for performing character recognition on a grayscale image in which a character string composed of characters arranged in a specific direction is displayed. The character recognition method, among the strings, N _C-number (N _C is 2 or more predetermined value) from the end includes the step of performing character recognition for the first character is a letter, N from the end of the string Performing character recognition for a second character other than _C characters. In character recognition for the first character, in character recognition for the first character, a plurality of first character areas of various character area sizes are set at respective positions in the grayscale image, and each of the plurality of first character areas is set. Character recognition processing and an evaluation value are calculated with respect to the character recognition result, and the character recognition result of the first character and each character position of the first character are determined and determined based on the calculated evaluation value. The optimum character spacing, which is the optimum character spacing, is determined from the determined character position. In character recognition for the second character, a plurality of second character areas are set at respective positions in the grayscale image in accordance with the optimum inter-character size, and character recognition processing is performed for each of the plurality of second character areas. The recognition result and the evaluation value are calculated, and the character recognition result of the second character is determined based on the character recognition result and the evaluation value calculated for each of the plurality of second character areas.

In still another aspect of the present invention, there is provided a program that causes an arithmetic device to perform character recognition on a grayscale image on which a character string composed of characters arranged in a specific direction is displayed. The program includes a step of performing character recognition for the first character that is N _C characters from the end (N _C is a predetermined value of 2 or more), and N _C characters from the end of the character string. And a step of performing character recognition on the second character that is a character other than the above. In character recognition for the first character, a plurality of first character regions having various character region sizes are set at respective positions in the gray scale image, and character recognition processing is performed for each of the plurality of first character regions. A result and an evaluation value are calculated, and based on the calculated evaluation value, a character recognition result of the first character and a first optimum character area size which is an optimum character area size are determined. In character recognition for the second character, a plurality of second character areas having a first optimum character area size are set at each position of the grayscale image, and character recognition processing is performed for each of the plurality of second character areas. The character recognition result and the evaluation value are calculated, and the character recognition result of the second character is determined based on the character recognition result and the evaluation value calculated for each of the plurality of second character areas.

In still another aspect of the present invention, there is provided a program that causes an arithmetic device to perform character recognition on a grayscale image on which a character string composed of characters arranged in a specific direction is displayed. The program includes a step of performing character recognition for the first character that is N _C characters from the end (N _C is a predetermined value of 2 or more), and N _C characters from the end of the character string. And a step of performing character recognition on the second character that is a character other than the above. In character recognition for the first character, in character recognition for the first character, a plurality of first character areas of various character area sizes are set at respective positions in the grayscale image, and each of the plurality of first character areas is set. Character recognition processing and an evaluation value are calculated with respect to the character recognition result, and the character recognition result of the first character and each character position of the first character are determined and determined based on the calculated evaluation value. The optimum character spacing, which is the optimum character spacing, is determined from the determined character position. In character recognition for the second character, a plurality of second character areas are set at respective positions in the grayscale image in accordance with the optimum inter-character size, and character recognition processing is performed for each of the plurality of second character areas. The recognition result and the evaluation value are calculated, and the character recognition result of the second character is determined based on the character recognition result and the evaluation value calculated for each of the plurality of second character areas.

  According to the present invention, when character recognition processing is performed on a grayscale image, it is possible to simultaneously improve the accuracy of character recognition and reduce the amount of calculation.

It is a block diagram which shows the structure of the character recognition apparatus in one Embodiment of this invention. It is a conceptual diagram which shows the example of the plate area | region image in which character recognition processing is performed in this embodiment. It is a flowchart which shows the outline | summary of the character recognition method of this embodiment. It is a figure which shows the definition of the character area set to a plate area image in this embodiment, and the position of this character area. It is a flowchart which shows the process about _NC character from one end in the character recognition method of this embodiment. It is a figure which shows notionally the content of the character recognition process about area | region position (i, j). It is a figure which shows notionally the content of the data processing performed about area | region size #k performed in step S15 of this embodiment. It is a figure which shows notionally the evaluation value-luminance value conversion plot image produced | generated in this embodiment. It is a list of average values μ _k of evaluation values obtained for a region size #k used in the character region size determination process of the present embodiment. It is a figure which shows the process which determines the character position and the size between characters in this embodiment. It is a conceptual diagram showing an example of the character recognition process performed for N _C +1 th character from one end.

  FIG. 1 is a block diagram showing a configuration of a character recognition device 1 according to an embodiment of the present invention. In the present embodiment, the character recognition device 1 is used for license plate recognition. In other words, the character recognition device 1 performs character recognition processing on the plate area image data 21 supplied from the outside to generate plate recognition data 22. Here, the plate area image data 21 is image data of a plate area image, that is, an image in which a license plate (automobile registration number mark or vehicle number mark) is projected. For example, from a captured image obtained by photographing a vehicle, It is obtained by image processing that cuts out a portion corresponding to the license plate. It should be noted that the plate area image data 21 is grayscale image data. The plate recognition data 22 is data indicating a character recognition result (that is, a recognized character) obtained by the character recognition process.

  The character recognition device 1 includes an image processing IC (integrated circuit) 2, an external interface 3, an external storage device 4, a memory 5, and a ROM (read only memory) 6. The external interface 3 supplies the plate area image data 21 received from the outside to the image processing IC 2. The external storage device 4 stores data generated in the character recognition process in the character recognition device 1. Data stored in the external storage device 4 includes plate recognition data 22. The memory 5 is used as a working area for arithmetic processing by the image processing IC 2. The ROM 6 stores a program executed by the image processing IC 2. The program stored in the ROM 6 includes character recognition software 6a that is a program for executing character recognition processing.

  The image processing IC 2 includes an arithmetic module 11, an image input interface 12, a data input / output interface 13, a memory controller 14, and a ROM controller 15. The arithmetic module 11, the image input interface 12, the data input / output interface 13, the memory controller 14, and the ROM controller 15 are connected by an internal bus 16. The arithmetic module 11 executes the character recognition software 6 a while using the memory 5 as a working area, and performs character recognition processing on the plate area image data 21. The image input interface 12 is an interface used for inputting the plate area image data 21 to the image processing IC 2. The data input / output interface 13 is an interface for accessing the external storage device 4. The memory controller 14 is an interface for accessing the memory 5. The ROM controller 15 is an interface for accessing the ROM 6.

  Each process of the character recognition method described below is executed using the hardware shown in FIG. Below, the character recognition method of this embodiment is demonstrated in detail.

  FIG. 2 illustrates an example of an image of plate area image data 21 on which character recognition processing is performed in the character recognition method of the present embodiment, that is, an example of a plate area image. In the present embodiment, the plate region image is a grayscale image, and is obtained by cutting out a portion corresponding to a license plate from a vehicle captured image obtained by photographing a vehicle. In the following description, description will be made using the xy coordinate system defined for the plate region image. Here, the x axis is defined in the character width direction (horizontal direction) of the plate area image, and the y axis is defined in the character height direction (vertical direction) of the plate area image. The character recognition method of the present embodiment is based on the premise that the size of each character is substantially the same in the plate area image. Such a premise is appropriate for license plate recognition, for example.

  FIG. 3 is a flowchart showing an outline of the character recognition method in this embodiment. The character recognition software 6a described above is a program code group for executing this character recognition method.

In the character recognition method of the present embodiment, first, character recognition processing is performed for N _C characters from one end (step S01). Here, N _C is the designated number of characters. The plate area image in FIG. 2 shows an example in which character recognition processing is performed on the leftmost three characters in step S01.

  In the character recognition process in step S01, character areas of various sizes are set at each position of the plate area image, and the character recognition process is performed for each of the set character areas. Here, the character area is an area where each character is considered to exist. FIG. 4 is a diagram illustrating an example of a character area. In the present embodiment, the character area 31 is a rectangle (that is, a rectangle (including a square)), and the position of each character area 31 is the center of the character area 31. That is, it is defined as the intersection 33 of the diagonal lines 32. Furthermore, based on the evaluation value calculated in the character recognition process, the optimum character recognition result (ie, the recognized character) and the optimum position and size of the character region are searched.

Subsequently, the character recognition process for the remaining characters is performed using the information obtained in the character recognition process for the N _C characters in step S01 (step S02). In step S02, information on the size of the character region and the character spacing obtained by the character recognition process in step S01 is used. The character recognition method of the present embodiment reduces the search range in the character recognition process in step S02 by such a method, and simultaneously improves the accuracy of character recognition and reduces the amount of calculation. Hereinafter, the process in each step will be described in detail.

FIG. 5 is a flowchart showing details of the character recognition processing for N _C characters in step S01. In the process in step S01, generally, after initial setting (step S11), character areas of various sizes are set at various positions, and character recognition processing is performed for each of the set character areas. (Steps S12 to S15). Furthermore, based on the evaluation value calculated in the character recognition process, the optimum character area size, character position, character spacing (character spacing), and optimum character recognition result are determined (steps S16 to S18). Hereinafter, the character recognition process for N _C characters in step S01 will be described in detail.

  In step S11, the character area initial size, the character area size increment (increase in the x direction and the y direction), the maximum value of the character area size (the maximum value in the x direction and the maximum value in the y direction), and the shift amount of the character area And are defined. The initial character area size is the size of the character area initially set. The size of the character area in the x direction (that is, the width of the character area) and the size in the y direction (that is, the height of the character area). Including. The character area size increment is an increment when the size of the character area is changed in the search for the optimum character area, and includes an x-direction increment and a y-direction increment. The maximum character area size is the maximum size when changing the size of the character area in the search for the optimal character area, and includes the maximum value in the x direction and the maximum value in the y direction. The shift amount of the character area is a designated value of the change amount of the character area when the position of the character area is changed in the search for the optimum character area, and includes the x-direction shift amount and the y-direction shift amount. . These initial settings are determined according to the position of the plate region image in the vehicle captured image.

  The processes in steps S12 to S15 are repeated as many times as the number of possible area sizes. First, the size of the character area is set (step S12). In the first process, the size of the character area is set to the character area initial size defined as the initial setting. On the other hand, in the second and subsequent processing, a size obtained by changing the initial size of the character region in accordance with the x-direction increment and the y-direction increment is selected as the size of the character region. The size of the character area set in step S12 is hereinafter referred to as area size #k. k is an index representing the region size.

  Further, the position of the character area is set (step S13). Hereinafter, the position set in step S13 is referred to as a region position (i, j). Here, i is an index representing a position in the x-axis direction, and j is an index representing a position in the y-axis direction. As described above, in this embodiment, the character area is rectangular, and the position of the character area is defined as the center of the character area. In the first process of step S13, the position of the character area is set to a predetermined process start position. In the plate region image, since the position where Nc characters exist from the left, in particular, the position of the leftmost character can be assumed within a certain range, the processing start position is the range of the assumed position. It may be determined appropriately according to the situation. In the second and subsequent processing, the position of the character area is set by changing the position of the character area in accordance with the shift amount of the character area included in the initial setting.

  Further, a character recognition process is performed for the character region whose position is determined in step S13 (step S14). When the position of the character area is determined to be the area position (i, j) in step S13, the character recognition result in the character area at the area position (i, j) in the character recognition process in step S14 (that is, recognized) Character) and its evaluation value. Furthermore, the character recognition results for the character region at the region position (i, j) are ranked based on the evaluation value. As a specific algorithm for the character recognition process in step S14, for example, an algorithm as disclosed in Non-Patent Document 1 described above may be used.

  FIG. 6 is a diagram showing a specific example of the character recognition process in step S14. Taking the character recognition processing of the plate area image of FIG. 2 as an example, for the leftmost character, for example, the character “S” with an evaluation value of 0.911 is obtained as the first character recognition result, and the evaluation value is The character “2” which is 0.566 is obtained as the second character recognition result. Here, the evaluation value is a value in the range of 0 to 1.0, and it is assumed that the higher the matching degree with the character template in the character recognition process, the closer to 1.0.

  Steps S13 and S14 are repeated while changing the position of the character area. The processes in steps S13 and S14 are repeated until the character recognition process is completed for all of the allowable character area positions. In the plate area image, a position where Nc characters are present from the left can be assumed to some extent. Therefore, the allowable range of the character area position may be appropriately determined according to the assumed position.

  When the character recognition processing is completed for all the allowable positions of the character area for the area size #k set in step S12, the data processing described below is performed for the character area with the higher evaluation value (step S12). S15).

In step S15, first, as shown in FIG. 7, for each region position (i, j), the character recognition result and evaluation value having the first rank are selected. Further, from the character recognition results of all the region positions having the first rank, m _C character recognition results and evaluation values having higher evaluation values are extracted. Further, an average value μ _k of the evaluation values of the extracted upper m _C character recognition results is calculated. Here, k is an index representing the region size, and μ _k represents an average value of the evaluation values of the upper m _C character recognition results calculated corresponding to the region size #k.

Further, from the evaluation values of the upper m _C (m _C is a predetermined value) character recognition results calculated corresponding to the region size #k, the evaluation value-luminance value conversion plot image shown in FIG. Is generated. Here, the evaluation value-luminance value conversion plot image is the character recognition result for each of the m _C character recognition results having the highest evaluation value among the character recognition results of all the region positions having the first rank. The pixel corresponding to the region position (i, j) is an image generated as having a luminance value corresponding to the evaluation value of the character recognition result. As a result, a non-zero luminance value is given for m _C pixels. The brightness value of other pixels is set to 0. The predetermined value m _C is a value that is appropriately determined to be larger than N _C , and a value that is several times (for example, five times) N _C is preset as the predetermined value m _C.

FIG. 8 shows an example of an evaluation value-luminance value conversion plot image. In FIG. 8, the evaluation value-luminance value conversion plot image corresponding to the region size #k is indicated by reference numeral 24-k. Here, n _S is the number of possible region sizes. As will be described later, the evaluation value-luminance value conversion plot image is used to determine the character position and the character size.

While updating the character area size, the processes in steps S12 to S15 are repeated. At this time, in step S12, the character area size is set while changing the character area size used in the immediately preceding process from the character area size based on the x-direction increment and the y-direction increment. Further, since step S15 is performed for each region size, as a result, as shown in FIG. 9, an average evaluation value μ _k is obtained for each region size #k.

After all the possible character area sizes are processed in steps S12 to S15, a character area size determination process is performed (step S16). In step S16, the highest corresponding area size to those of the average value mu _k calculated in step S15 is finally employed as the character area size. As will be described later, the character area size employed in step S16 is used for character recognition processing for the N _C +1 and subsequent characters.

  Subsequently, the character position of each character and the inter-character size are determined (step S17). Here, in the present embodiment, the character position of each character is defined as the center of the character area corresponding to the character, and the inter-character size is defined as an interval between character positions of adjacent characters.

  FIG. 10 is a conceptual diagram showing a procedure for determining the character position and the inter-character size in step S17. First, an evaluation value-luminance value conversion plot image corresponding to the character area size employed in step S16 is selected (step S21). When the character area size adopted in step S16 is the area size #k, the evaluation value-luminance value conversion plot image 24-k is selected.

  Further, the selected evaluation value-luminance value conversion plot image is subjected to a binarization process for binarization with a predetermined luminance threshold, and an image obtained by the binarization process (binarized image). Is subjected to a labeling process (step S22). The labeling process is a process for assigning the same label to consecutive pixels of the same color (white or black) for the binarized image. In the present embodiment, in the evaluation value-luminance value conversion plot image, binarization processing is performed assuming that pixels whose luminance value is higher than the luminance threshold value are “white” pixels, and other pixels are “black” pixels. In addition, the same label is given to a set of consecutive “white” pixels. FIG. 10 illustrates an example in which four labels # 1 to # 4 are given in step S22.

If the number of labels determined in step S22 is greater than the number of characters to be subjected to character recognition processing in step S01 (that is, the designated number of characters N _C ), labels that are inappropriate as character positions are eliminated ( Step S23). In step S23, firstly, for each label #L, average mu _L of the evaluation value is calculated. The luminance value of each pixel belonging to each label corresponds to the evaluation value of the character recognition result of the character area at the position of the pixel, and the average value of the evaluation values corresponding to the pixels belonging to the label #L is μ _L Calculated. Further, the barycentric position of each label #L is calculated. Since each label #L is composed of one or a plurality of pixels, the barycentric position of each label #L is calculated from the positions of the pixels belonging to the label #L.

Furthermore, in order from the higher average value mu _L label, calculates the distance between the gravity center position between the position of the center of gravity and the other labels in each label, the distance between the center of gravity of the other labels is equal to or less than a predetermined threshold D _TH process to eliminate the label if it has a center of gravity position such that is repeatedly performed until the same number of labels and specified number N _C is left. By this processing, each label is associated with each of the N _C characters from the left end.

  Further, among the pixels of each label, the pixel having the highest evaluation value of the corresponding character recognition result is selected, and the position of the pixel of the character recognition result having the highest evaluation value is determined as the character position of each character (step) S24). Here, it should be noted that in this embodiment, the character position of each character is defined as the center position of the corresponding character area.

Further, the distance between adjacent character positions in the x-axis direction is calculated, and the inter-character size (that is, the character spacing) is determined based on the calculated distance. When there are a plurality of distances in the x-axis direction between adjacent character positions, the inter-character size is determined by calculating the plurality of distances. For example, the calculated average value of the plurality of distances may be used as the inter-character size. As will be described later, the inter-character size determined in step S24 is used for the character recognition processing for the N _C +1 and subsequent characters.

  Furthermore, the character recognition result for the character region located at the character position of each character determined in step S17 and having the character region size determined in step S16 is selected as the character recognition result for each character ( Step S18).

By the above procedure, along with the character recognition results of N _C character from the left side is obtained, the optimal character area size (i.e., character region size determined in step S16), and and, among the optimum character size (i.e., step S17 The inter-character size determined in (1) is obtained.

Step S02:
Subsequently, character recognition processing is performed on the remaining characters (in this embodiment, the characters after the N _C +1 character from the left). In the character recognition processing for the remaining characters, the optimum character area size and inter-character size obtained in the above-described processing are used, thereby reducing the search range in the character recognition processing.

FIG. 11 is a diagram for explaining the character recognition process of the N _C +1 character. In the process of the N _C +1 character, the character position (determined in step S17 described above) shifted from the character position of the N _C character in the x-axis direction by the optimum inter-character size D _X obtained in the above process. As the position, the reference position of the character position of the N _C +1 character is determined. Further, the character area is set while sequentially changing the position between the reference position and the surrounding area, and character recognition processing is performed for each set character area. At this time, as the reference size (width and height) of the character area to be set, the optimum character area size obtained by the above processing is used. In FIG. 11, among the optimum character region sizes, the size in the x direction (ie, width) is indicated by the symbol “W”, and the size in the y direction (ie, height) is indicated by the symbol “H”. ing. The width and height of the character area set in the character recognition process are set while sequentially changing the width and height of the reference size and the surrounding area. Further, the character recognition result (that is, the recognized character) having the highest evaluation value calculated in the character recognition process and the position of the character area are determined as the character recognition result and the character position of the N _C +1 character.

Furthermore, the same processing is performed for the characters after the N _C + 2nd character from the left side. The reference position of the character position of the N _C +2 character is determined at a position shifted in the x-axis direction by the inter-character size D _X from the character position of the N _C +1 character (determined by the above-described processing). . Further, the character area is sequentially changed by changing the position of the character area between the reference position and the surrounding area, and the width and height of the character area are changed to the width and height of the reference size and the surrounding area. The character recognition process is performed for each set character area. Further, the character recognition result (that is, the recognized character) and the character area position having the highest evaluation value calculated in the character recognition process are determined as the character recognition result and the character position of the N _C + 2nd character. Thereafter, the same processing is repeated until the character recognition processing is completed for all characters.

Here, when it is assumed that there are a plurality of types of license plates displayed in the plate region image, the type of the license plate is specified by the information on the inter-character size obtained by the above processing, and N _C In the character recognition processing for the + 1st character and after, processing corresponding to the specified type of license plate may be performed. For example, a countermeasure process for a specific similar character may be performed.

In the character recognition method according to the present embodiment described above, the character recognition process is performed without performing the binarization process on the plate area image. For this reason, the problem accompanying a binarization process can be avoided. For example, character recognition is possible even if the plate area image has a non-uniform luminance portion. In addition, in the character recognition method of the present embodiment, the character recognition process for the remaining characters is performed using the information on the character area size and the inter-character size obtained in the character recognition process for the _NC character from the end. The search range in the character recognition process for the remaining characters can be reduced.

In the above embodiment, the character recognition process for the remaining characters is performed using both the character area size and the inter-character size information obtained in the character recognition process for the _NC character from the end. The character recognition processing for the remaining characters may be performed using only one of the size and the size between characters. Even in this case, the effect of reducing the search range can be obtained.

  Further, in the above-described embodiment, the character recognition processing is performed in order from the left character, but the order may be reversed. For example, character recognition processing may be performed in order from the right character. When the characters are arranged in the vertical direction, the character recognition process may be performed in order from the upper character, and conversely, the character recognition process may be performed in order from the lower character.

  Furthermore, the character recognition processing described above may be performed twice in the order opposite to each other, and the character recognition results obtained by the character recognition processing may be compared. In this case, when a character recognition result of a certain character is different, a character recognition result having a high evaluation value may be selected as a final character recognition result of the character.

In a modification, first, as in the above-described embodiment, the character recognition process and the evaluation value of each character are obtained by performing the character recognition processing in order from the left character. Furthermore, the character recognition process and the evaluation value of each character are obtained by performing the character recognition processing in order from the right character. More specifically, first, the right N _C characters by performing the same process as in step S01, after obtaining the character recognition result and the evaluation value of the N _C characters, similarly to step S02 for the remaining characters The character recognition result and evaluation value of the remaining characters are obtained. In this case, for the remaining characters, a character recognition result and an evaluation value are obtained in order from the right character. Thereafter, the character recognition result obtained by performing the character recognition processing in order from the left character is compared with the character recognition result obtained by performing the character recognition processing in order from the right character. For characters having the same character recognition result, the same character recognition result is selected as the final character recognition result of the character. On the other hand, for characters having different character recognition results, a character recognition result having a high evaluation value is selected as the final character recognition result of the character.

  According to such a method, when shadows or the like are generated on a part of the plate, the character area size and the character interval size are determined based on the result of the area without the shadow (resulting in the high evaluation value). Therefore, the recognition accuracy can be improved.

  As another modification of the present embodiment, when a character (similar character) that is likely to be erroneously recognized is known, when the similar character is obtained as the character recognition result, the character recognition result and the evaluation value The character area size may be determined without using. Examples of similar characters include B and 8, D and 0, 7 and Z, and the like.

For example, in step S15, the character recognition result and the evaluation value that are ranked first for each region position (i, j) are selected, and further, from among the character recognition results of all the region positions that are ranked first. If the evaluation value is carried out a process of extracting m _C-number of the character recognition result of the upper and evaluation value, the character recognition result and the evaluation values of m _C-number except for character recognition result is similar characters may be extracted . In this case, the average value μ _k is calculated as the average value of the evaluation values of the upper m _C character recognition results, excluding the character recognition results of similar characters. At this time, the list of similar characters may be stored in an appropriate storage unit (for example, the external storage device 4).

  Although the embodiment of the present invention has been specifically described above, the present invention is not limited to the above embodiment. It will be apparent to those skilled in the art that the present invention can be implemented with various modifications. For example, in the above description, an embodiment in which the present invention is applied to license plate recognition (recognition of characters written on a license plate) is described. However, the present invention generally recognizes character recognition for captured images that are grayscale images. Applicable.

1: Character recognition device 2: Image processing IC
3: External interface 4: External storage device 5: Memory 6: ROM
6a: Character recognition software 11: Arithmetic module 12: Image input interface 13: Data input / output interface 14: Memory controller 15: ROM controller 16: Internal bus 21: Plate area image data 22: Plate recognition data 31: Character area 32: Diagonal line 33: Intersection

In one aspect of the present invention, a character recognition device is provided that performs character recognition on a grayscale image in which a character string composed of characters arranged in a specific direction is displayed. Character recognition apparatus, among the strings, N _C-number from the first end (N _C is 2 or more predetermined value) for the first character is a character of the first character recognition means for performing character recognition, the character string And second character recognition means for performing character recognition on a second character which is a character other than N _C characters from the first end. In the character recognition for the first character, the first character recognition means sets a plurality of first character regions having various character region sizes at respective positions in the gray scale image, and performs character recognition for each of the plurality of first character regions. Processing is performed to calculate a character recognition result and an evaluation value, and based on the calculated evaluation value, a character recognition result of the first character and a first optimum character area size that is an optimum character area size are determined. To do. The second character recognition means sets a plurality of second character areas having the first optimum character area size at each position of the gray scale image, performs character recognition processing for each of the plurality of second character areas, and performs a character recognition result. And the evaluation value, and the character recognition result of the second character is determined based on the character recognition result and the evaluation value calculated for each of the plurality of second character regions.

In this case, for each character area size, the first character recognizing means includes pixels having luminance values corresponding to the evaluation values of m _C character recognition results, corresponding to the m _C character recognition results. An evaluation value-luminance value conversion plot image, which is an image located at a region position, is generated and a binarization process is performed on the evaluation value-luminance value conversion plot image to generate a binarized image. preferable. The first character recognition means performs a labeling process on the binarized image to give a label to each set of continuous pixels having the same color, and for each label, character recognition corresponding to a pixel belonging to the label A second average value that is an average value of the evaluation values of the results is calculated, and the first character recognizing unit sequentially calculates the centroid position of each label and the centroid position of other labels in order from the label having the second highest average value. The process of removing each label is repeated until N _C labels are left when the distance between the labels is calculated and the center of gravity has a center of gravity such that the distance from the center of gravity of another label is equal to or less than a predetermined threshold. Do. The first character recognition means, for each of the N _C-number of labels remaining, among the pixels belonging to each selects the highest pixel evaluation value of the corresponding character recognition result is the highest pixel evaluation value Is determined as the character position of the first character, and the optimal inter-character size, which is the optimal character spacing, is determined from the determined character position.
The second character recognition means sets the positions of the plurality of second character areas according to the optimum inter-character size.

In one embodiment, the character recognition device, further in the character string, and a third character recognition means for performing character recognition for the third character is a N _C-number of characters from the opposite side of the second end and the first end, A fourth character recognition unit that performs character recognition on a fourth character that is a character other than N _C characters from the second end of the character string, and a final character recognition result determination unit may be included. In the character recognition for the third character, the third character recognition means sets a plurality of third character areas having various character area sizes at respective positions in the gray scale image, and performs character recognition for each of the plurality of third character areas. Processing is performed to calculate a character recognition result and an evaluation value, and based on the calculated evaluation value, a character recognition result for the third character and a second optimum character area size that is an optimum character area size are determined. To do. The fourth character recognition means sets each position of the gray scale image of the plurality of fourth character areas having the second optimum character area size, performs character recognition processing for each of the plurality of fourth character areas, and performs a character recognition result. And the evaluation value, and the character recognition result of the fourth character is determined based on the calculated evaluation value. The final character recognition result determining means is a method in which the character recognition result of the specific character of the character string obtained by the first character recognition means and the second character recognition means is obtained by the third character recognition means and the fourth character recognition means. When the character recognition result matches the character recognition result, the matching character recognition result is determined as the final character recognition result of the specific character. If they do not match, the final character recognition result determination means is obtained by the character recognition result of the specific character obtained by the first character recognition means and the second character recognition means, and by the third character recognition means and the fourth character recognition means. The character recognition result having the higher evaluation value among the character recognition results of the specific character is determined as the final character recognition result of the specific character.

In another aspect of the present invention, there is provided a character recognition device that performs character recognition on a grayscale image on which a character string composed of characters arranged in a specific direction is displayed. The character recognition apparatus, in the character string, N _C-number (N _C is 2 or more predetermined value) from the end and the first character recognition means for performing character recognition for the first character is a character, an end character string To a second character recognition means for performing character recognition on the second character which is a character other than N _C characters. In the character recognition for the first character, the first character recognition means sets a plurality of first character regions having various character region sizes at respective positions in the gray scale image, and performs character recognition for each of the plurality of first character regions. A character recognition result and an evaluation value are calculated by performing processing, and the character recognition result of the first character and each character position of the first character are determined based on the calculated evaluation value, and the determined character position To determine the optimum character spacing, which is the optimum character spacing. The second character recognition means sets a plurality of second character areas at each position of the grayscale image according to the optimum inter-character size, performs character recognition processing for each of the plurality of second character areas, The evaluation value is calculated, and the character recognition result of the second character is determined based on the character recognition result and the evaluation value calculated for each of the plurality of second character regions.

In still another aspect of the present invention, there is provided a character recognition method for performing character recognition on a grayscale image in which a character string composed of characters arranged in a specific direction is displayed. The character recognition method, in the character string, N _C-number (N _C is 2 or more predetermined value) from the end and performing character recognition for the first character is a character of, N _C-number from the end of the string Recognizing a second character that is not a character. In character recognition for the first character, a plurality of first character regions having various character region sizes are set at respective positions in the gray scale image, and character recognition processing is performed for each of the plurality of first character regions. A result and an evaluation value are calculated, and based on the calculated evaluation value, a character recognition result of the first character and a first optimum character area size which is an optimum character area size are determined. In character recognition for the second character, a plurality of second character areas having a first optimum character area size are set at each position of the grayscale image, and character recognition processing is performed for each of the plurality of second character areas. The character recognition result and the evaluation value are calculated, and the character recognition result of the second character is determined based on the character recognition result and the evaluation value calculated for each of the plurality of second character areas.

In still another aspect of the present invention, there is provided a character recognition method for performing character recognition on a grayscale image in which a character string composed of characters arranged in a specific direction is displayed. The character recognition method, in the character string, N _C-number (N _C is 2 or more predetermined value) from the end and performing character recognition for the first character is a character of, N _C-number from the end of the string Recognizing a second character that is not a character. In character recognition for the first character, in character recognition for the first character, a plurality of first character areas of various character area sizes are set at respective positions in the grayscale image, and each of the plurality of first character areas is set. Character recognition processing and an evaluation value are calculated with respect to the character recognition result, and the character recognition result of the first character and each character position of the first character are determined and determined based on the calculated evaluation value. The optimum character spacing, which is the optimum character spacing, is determined from the determined character position. In character recognition for the second character, a plurality of second character areas are set at respective positions in the grayscale image in accordance with the optimum inter-character size, and character recognition processing is performed for each of the plurality of second character areas. The recognition result and the evaluation value are calculated, and the character recognition result of the second character is determined based on the character recognition result and the evaluation value calculated for each of the plurality of second character areas.

In still another aspect of the present invention, there is provided a program that causes an arithmetic device to perform character recognition on a grayscale image on which a character string composed of characters arranged in a specific direction is displayed. The program, in the character string, N _C-number (N _C is 2 or more predetermined value) from the end and performing character recognition for the first character is a character, from the end of the character string other than N _C atoms And causing the arithmetic unit to execute character recognition for the second character, which is a character. In character recognition for the first character, a plurality of first character regions having various character region sizes are set at respective positions in the gray scale image, and character recognition processing is performed for each of the plurality of first character regions. A result and an evaluation value are calculated, and based on the calculated evaluation value, a character recognition result of the first character and a first optimum character area size which is an optimum character area size are determined. In character recognition for the second character, a plurality of second character areas having a first optimum character area size are set at each position of the grayscale image, and character recognition processing is performed for each of the plurality of second character areas. The character recognition result and the evaluation value are calculated, and the character recognition result of the second character is determined based on the character recognition result and the evaluation value calculated for each of the plurality of second character areas.

In still another aspect of the present invention, there is provided a program that causes an arithmetic device to perform character recognition on a grayscale image on which a character string composed of characters arranged in a specific direction is displayed. The program, in the character string, N _C-number (N _C is 2 or more predetermined value) from the end and performing character recognition for the first character is a character, from the end of the character string other than N _C atoms And causing the arithmetic unit to execute character recognition for the second character, which is a character. In character recognition for the first character, in character recognition for the first character, a plurality of first character areas of various character area sizes are set at respective positions in the grayscale image, and each of the plurality of first character areas is set. Character recognition processing and an evaluation value are calculated with respect to the character recognition result, and the character recognition result of the first character and each character position of the first character are determined and determined based on the calculated evaluation value. The optimum character spacing, which is the optimum character spacing, is determined from the determined character position. In character recognition for the second character, a plurality of second character areas are set at respective positions in the grayscale image in accordance with the optimum inter-character size, and character recognition processing is performed for each of the plurality of second character areas. The recognition result and the evaluation value are calculated, and the character recognition result of the second character is determined based on the character recognition result and the evaluation value calculated for each of the plurality of second character areas.

Claims (11)

A character recognition device that performs character recognition on a grayscale image in which a character string composed of characters arranged in a specific direction is displayed,
Among the character strings, N _C-number (N _C is 2 or more predetermined value) from the first end includes a first character recognition means for performing character recognition for the first character is a character,
Second character recognition means for performing character recognition for a second character that is a character other than the first character,
In the character recognition for the first character, the first character recognition means sets a plurality of first character areas of various character area sizes at respective positions of the gray scale image, and sets the first character areas of the plurality of first character areas. Character recognition processing and evaluation values are calculated by performing character recognition processing for each, and the character recognition result of the first character is determined based on the calculated evaluation values, and among the various character area sizes 1st optimal character area size which is the size of the optimal character area is determined from
The second character recognition means sets a plurality of second character regions having the first optimum character region size at each position of the gray scale image, and performs character recognition processing for each of the plurality of second character regions. The character recognition result and the evaluation value are calculated, and the character recognition result of the second character is determined based on the character recognition result and the evaluation value calculated for each of the plurality of second character regions. Character recognition device. The character recognition device according to claim 1,
The first character recognition means determines each character position of the first character based on the evaluation value, and determines an optimal character size that is an optimal character spacing from the determined character position,
The character recognition device, wherein the second character recognition means sets the positions of the plurality of second character areas according to the optimum inter-character size. The character recognition device according to claim 1,
The first character recognition means ranks the character recognition results obtained for each position of the plurality of first character areas based on the evaluation value;
The first character recognizing means has an evaluation value from among character recognition results obtained for each position of the plurality of first character regions and ranked first for each of the character region sizes. Extracting high m _C pieces ( _mC is a predetermined number larger than N _C ) and calculating a first average value that is an average value of evaluation values of the m _C character recognition results;
The first character recognizing means determines a character area size having the highest first average value as the optimum character area size. The character recognition device according to claim 3,
First the character recognition means, for each of the character area size, pixels having a luminance value corresponding to the evaluation value of the m _C-number of the character recognition result corresponds to each of the m _C-number of the character recognition result An evaluation value-luminance value conversion plot image, which is an image located at the position of the character area, is generated, and a binarization process is performed on the evaluation value-luminance value conversion plot image to generate a binary image. ,
The first character recognition means performs a labeling process on the binarized image to give a label to each set of continuous pixels having the same color, and assigns a label to each pixel belonging to the label. Calculating a second average value that is an average value of the evaluation values of the corresponding character recognition results;
The first character recognition means calculates the distance between each barycentric position of each label and the barycentric position of the other label in order from the label having the second average higher value, and the barycentric position of the other label. When the center of gravity position is such that the distance is equal to or less than a predetermined threshold, the process of removing each label is repeated until N _C labels are left,
The first character recognizing means selects, for each of the remaining N _C labels, a pixel having the highest evaluation value of the corresponding character recognition result as a pixel belonging to each of the N _C labels, and having the highest evaluation value A pixel position is determined as the character position of the first character, and an optimal character size that is an optimal character spacing is determined from the determined character position;
The character recognition device, wherein the second character recognition means sets the positions of the plurality of second character areas according to the optimum inter-character size. The character recognition device according to any one of claims 1 to 4,
Among the character string, the N _C-number from the second end opposite the first end, and a third character recognition means for performing character recognition for the third character is a character,
N _C-number from the second end of the string, and the fourth character recognition means for performing character recognition fourth character is a character other than the characters,
A final character recognition result determining means;
In the character recognition for the third character, the third character recognition means sets a plurality of third character regions having various character region sizes at respective positions in the gray scale image, and Character recognition processing and an evaluation value are calculated by performing character recognition processing for each, and based on the calculated evaluation value, the character recognition result of the third character and the second optimum that is the size of the optimum character area Determine the character area size,
The fourth character recognition means sets each position of the plurality of fourth character areas having the second optimum character area size in the gray scale image, and performs character recognition processing for each of the plurality of fourth character areas. Calculating the character recognition result and the evaluation value, and determining the character recognition result of the fourth character based on the calculated evaluation value,
The final character recognition result determination means is configured to obtain a character recognition result of a specific character of the character string obtained by the first character recognition means and the second character recognition means as the third character recognition means and the fourth character recognition. If the character recognition result of the specific character obtained by the means matches, the matching character recognition result is determined as the final character recognition result of the specific character, and if the character recognition result does not match, the first character recognition means and the One having a higher evaluation value among the character recognition result of the specific character obtained by the second character recognition unit and the character recognition result of the specific character obtained by the third character recognition unit and the fourth character recognition unit A character recognition apparatus that determines a character recognition result of the character recognition result as a final character recognition result of the specific character. The character recognition device according to any one of claims 1 to 4,
The first character recognizing means includes storage means for storing similar character data in which similar characters are registered,
The first character recognizing means uses the character recognition result that matches the similar character among the character recognition results obtained by performing the character recognition processing for each of the plurality of first character regions and the evaluation value without using the character recognition result and the evaluation value thereof. A character recognition device that determines the optimal character area size. A character recognition device that performs character recognition on a grayscale image in which a character string composed of characters arranged in a specific direction is displayed,
Among the character strings, N _C-number from the end (N _C is 2 or more predetermined value), a first character recognition means for performing character recognition for the first character is a character,
Second character recognition means for performing character recognition for a second character that is a character other than the first character,
In the character recognition for the first character, the first character recognition means sets a plurality of first character areas of various character area sizes at respective positions of the gray scale image, and sets the first character areas of the plurality of first character areas. Character recognition processing and an evaluation value are calculated for each of them, and a character recognition result of the first character and a character position of the first character are determined based on the calculated evaluation value. And determining the optimum inter-character size that is the optimum character spacing from the determined character position,
The second character recognition means sets a plurality of second character areas at each position of the gray scale image according to the optimum inter-character size, and performs character recognition processing for each of the plurality of second character areas. A character that calculates a character recognition result and an evaluation value, and determines a character recognition result of the second character based on the character recognition result and the evaluation value calculated for each of the plurality of second character regions Recognition device. A character recognition method for performing character recognition on a grayscale image in which a character string composed of characters arranged in a specific direction is displayed,
Among the character strings, N _C-number from the end (N _C is 2 or more predetermined value), the step of performing character recognition for the first character is a character,
Performing character recognition for a second character that is a character other than the first character,
In character recognition for the first character, a plurality of first character regions having various character region sizes are set at respective positions in the gray scale image, and character recognition processing is performed for each of the plurality of first character regions. And calculating a character recognition result and an evaluation value, determining a character recognition result of the first character based on the calculated evaluation value, and selecting an optimum character area size from the various character area sizes. And determine the first optimum character area size,
In character recognition for the second character, a plurality of second character areas having the first optimum character area size are set at respective positions in the grayscale image, and character recognition is performed for each of the plurality of second character areas. A character recognition result and an evaluation value are calculated by performing processing, and a character recognition result of the second character is calculated based on the character recognition result and the evaluation value calculated for each of the plurality of second character regions. Determine the character recognition method. A character recognition method for performing character recognition on a grayscale image in which a character string composed of characters arranged in a specific direction is displayed,
Among the character strings, N _C-number from the end (N _C is 2 or more predetermined value), the step of performing character recognition for the first character is a character,
Performing character recognition for a second character that is a character other than the first character,
In character recognition for the first character, in character recognition for the first character, a plurality of first character regions of various character region sizes are set at each position of the grayscale image, and the plurality of first characters Character recognition processing is performed for each character region to calculate a character recognition result and an evaluation value, and based on the calculated evaluation value, the character recognition result of the first character and the character position of the first character And determining an optimal inter-character size that is an optimal character spacing from the determined character position,
In character recognition for the second character, a plurality of second character areas are set at each position of the gray scale image according to the optimum inter-character size, and character recognition processing is performed for each of the plurality of second character areas. The character recognition result and the evaluation value are calculated, and the character recognition result of the second character is calculated based on the character recognition result and the evaluation value calculated for each of the plurality of second character regions. Determine the character recognition method. A program that causes a computing device to perform character recognition on a grayscale image in which a character string composed of characters arranged in a specific direction is displayed,
Among the character strings, N _C-number from the end (N _C is 2 or more predetermined value), the step of performing character recognition for the first character is a character,
Performing the character recognition on the second character, which is a character other than the first character,
In character recognition for the first character, a plurality of first character regions having various character region sizes are set at respective positions in the gray scale image, and character recognition processing is performed for each of the plurality of first character regions. And calculating a character recognition result and an evaluation value, determining a character recognition result of the first character based on the calculated evaluation value, and selecting an optimum character area size from the various character area sizes. And determine the first optimum character area size,
In character recognition for the second character, a plurality of second character areas having the first optimum character area size are set at respective positions in the grayscale image, and character recognition is performed for each of the plurality of second character areas. A character recognition result and an evaluation value are calculated by performing processing, and a character recognition result of the second character is calculated based on the character recognition result and the evaluation value calculated for each of the plurality of second character regions. Determine the program. A program that causes a computing device to perform character recognition on a grayscale image in which a character string composed of characters arranged in a specific direction is displayed,
Among the character strings, N _C-number from the end (N _C is 2 or more predetermined value), the step of performing character recognition for the first character is a character,
Performing the character recognition on the second character, which is a character other than the first character,
In character recognition for the first character, in character recognition for the first character, a plurality of first character regions of various character region sizes are set at each position of the grayscale image, and the plurality of first characters Character recognition processing is performed for each character region to calculate a character recognition result and an evaluation value, and based on the calculated evaluation value, the character recognition result of the first character and the character position of the first character And determining an optimal inter-character size that is an optimal character spacing from the determined character position,
In character recognition for the second character, a plurality of second character areas are set at each position of the gray scale image according to the optimum inter-character size, and character recognition processing is performed for each of the plurality of second character areas. The character recognition result and the evaluation value are calculated, and the character recognition result of the second character is calculated based on the character recognition result and the evaluation value calculated for each of the plurality of second character regions. Deciding program.

Images