KR102208683B1 - Character recognition method and apparatus thereof - Google Patents

Abstract

A method and apparatus for recognizing characters from low-quality input images containing characters are provided. The character recognition method according to an embodiment of the present invention includes the steps of extracting a feature value from an input image, and determining an applied font for a character displayed in the input image by using the feature value and font shape information of each font. , Using the font feature information of the applied font, dividing the input image into each character area image, comparing the character area image with each character template of the applied font, and each character template of the character area image And calculating a degree of similarity with respect to, and determining a character corresponding to the character area image using the degree of similarity.

Description

Character recognition method and apparatus thereof TECHNICAL FIELD

The present invention relates to a character recognition method and apparatus thereof. In more detail, the present invention relates to a method for recognizing an input image of low quality containing characters, that is, an input image having an atypical form in which the character part is contaminated by external factors or the character composition is uneven, and the apparatus to which the method is applied. .

Character recognition methods can be broadly classified into pattern matching and structure analysis. The former is mainly used for the recognition of printed characters, and the latter is mainly used for the recognition of handwritten characters. The intermediate character recognition methods include feature matching and stroke analysis, and are used separately for each purpose or purpose.

On the other hand, when it is necessary to recognize characters displayed on the surface of materials such as steel and wood, the conventional character recognition method is difficult to apply. This is because the surface of the material is uneven, unlike paper, or because of the nature of the process, there are many variables that cause the letters to be corroded, contaminated or damaged. In addition, the form of the character is diverse as the number of dots (打刻) characters are also occupied. Therefore, there is a need to provide a method of recognizing characters from images with poor character display status, such as images including characters displayed on the surface of materials.

Korean Patent Publication No. 2014-0007764

The technical problem to be solved by the present invention is to provide a method and an apparatus for recognizing characters from an input image of low quality including characters.

Another technical problem to be solved by the present invention is to determine the font of the character using the feature value of the input image containing the character, divide the input image into each character area image, and then determine the character area image. It is to provide a method and apparatus for comparing each character template included in a character template and recognizing characters using the comparison result.

Another technical problem to be solved by the present invention is to provide a method and an apparatus for recognizing a character even when a character in an image is displayed in a dot type.

The technical problems of the present invention are not limited to the technical problems mentioned above, and other technical problems that are not mentioned will be clearly understood by those skilled in the art from the following description.

In the data verification method according to an embodiment of the present invention for achieving the above technical problem, the step of extracting a feature value from an input image, using the feature value and font shape information of each font, Determining an applied font for the applied font, dividing the input image into each character area image using font feature information of the applied font, comparing the character area image with each character template of the applied font, And calculating a similarity for each character template of the character area image, and determining a character corresponding to the character area image by using the similarity level.

The extracting of the feature value may include applying a predetermined correction routine to the input image to extract a high quality edge from the input image, and extracting an edge from the input image. It may include.

The determining of the applied font includes determining the size of each character displayed in the input image and the thickness of a stroke constituting the character using the extracted edge, and conforming to the size and thickness of the character. It may include the step of determining a font having the font shape information as the applied font.

The determining of the applied font includes determining as a dot type font when there are more than a predetermined number of circular edges among the extracted edges, grouping adjacent circular edges, and each circle belonging to one group. Determining the size of each character by using the coordinates of the pixels constituting the edge, determining the thickness of the dot using the diameter of each circular edge, and the font shape corresponding to the size of the character and the thickness of the dot It may include the step of determining a dot type font having information as the applied font.

The calculating of the similarity may include matching the size of the character area image with the size of the image of each character template of the applied font.

The calculating of the similarity may include calculating a similarity by calculating a pixel value matching ratio between the character area image and the image of each character template of the applied font, and the character area image is equivalent to each character template of the applied font. It may include the step of adjusting the degree of similarity depending on whether or not the characteristic requirement of is satisfied.

The separating may include setting a character separation guideline in the input image using a character spacing, a long pyeong, a font image width, and a font image height among font characteristic information of the applied font, and based on the character separation guideline. It may include the step of dividing the input image into each character area image.

The data verification method according to another embodiment of the present invention for achieving the above technical problem is the step of dividing an input image into each character area image by using font characteristic information, and comparing the character area image with each character template. And calculating a similarity degree for each character template of the character area image, and determining a character corresponding to the character area image by using the similarity level. In this case, calculating the degree of similarity includes reflecting whether the character area image satisfies the characteristic requirements of each character template.

A character recognition apparatus according to another embodiment of the present invention for achieving the above technical problem includes an image feature extracting unit for extracting a feature value from an input image, the input image using the feature value and font shape information of each font. A font determination unit that determines an applied font for the characters displayed in, a character area partition unit that divides the input image into each character area image using font characteristic information of the applied font, and the character area image. And a character determination unit for calculating a similarity of each character template of the character area image by comparing it with each character template of the applied font, and determining a character corresponding to the character area image by using the similarity level.

According to an embodiment, the character recognition apparatus performs a predetermined image quality correction image processing on the character area image, performs binarization of the character area image on which the image processing is completed, and the binarized character area image It may further include a character determination unit providing the character determination unit.

According to the present invention as described above, there is an effect of recognizing a character from an image having a poor character display state, such as an image including characters displayed on the surface of a material. In particular, there is an effect of being able to recognize effectively even when contamination exists in a character display portion or an irregular character with a non-uniform composition.

In addition, there is an effect of recognizing characters from an image containing characters displayed in a dot type.

1 is a view for explaining the effect of the present invention.
2 is a flowchart of a character recognition method according to an embodiment of the present invention.
3 is a diagram illustrating a result of image classification performed in some steps of the character recognition method shown in FIG. 2.
FIG. 4 is a detailed flowchart for explaining in detail some steps of the method for recognizing characters illustrated in FIG. 2.
FIG. 5 is a conceptual diagram illustrating some steps in the character recognition method shown in FIG. 2.
6 is an example of a user interface displayed as a result of performing the method described with reference to FIGS. 2 to 5.
7 is a diagram showing a character recognition apparatus according to another embodiment of the present invention.
8 is a block diagram of a character recognition apparatus according to another embodiment of the present invention.
9 is a detailed block diagram of some blocks of the character recognition apparatus shown in FIG. 8.
10 is a diagram showing a hardware configuration of a character recognition apparatus according to another embodiment of the present invention.

Hereinafter, preferred embodiments of the present invention will be described in detail with reference to the accompanying drawings. Advantages and features of the present invention, and a method of achieving them will become apparent with reference to the embodiments described below in detail together with the accompanying drawings. However, the present invention is not limited to the embodiments to be posted below, but may be implemented in a variety of different forms, and only these embodiments make the posting of the present invention complete, and common knowledge in the technical field to which the present invention pertains. It is provided to completely inform the scope of the invention to those who have it, and the invention is only defined by the scope of the claims. The same reference numerals refer to the same components throughout the specification.

Unless otherwise defined, all terms (including technical and scientific terms) used in the present specification may be used as meanings that can be commonly understood by those of ordinary skill in the art to which the present invention belongs. In addition, terms defined in a commonly used dictionary are not interpreted ideally or excessively unless explicitly defined specifically. The terms used in the present specification are for describing exemplary embodiments and are not intended to limit the present invention. In this specification, the singular form also includes the plural form unless specifically stated in the phrase. As used in the specification, "comprises" and/or "comprising" refers to the presence of one or more other components, steps, actions and/or elements, and/or elements, steps, actions and/or elements mentioned. Or does not exclude additions.

In this specification, a logic (LOGIC) or a routine (ROUTINE) refers to a series of instructions (OPERATION) that can be executed by a processor, and is not limited to being written by a specific programming language.

First, a character recognition method according to embodiments of the present invention and an image including characters that can be recognized by the apparatus and a recognition result thereof are presented with reference to FIG. 1.

As already mentioned, related information, such as the serial number of the material, can be stamped on the surface of a material such as wood, plate, or steel. The character to be struck may be a general character composed of connected strokes, but may be a dot type character for efficiency of struck. In addition, depending on various situations occurring in the process, the stamped characters may have an atypical shape. In addition, the surface of the material may be uneven, unlike paper, or the character part may be corroded, contaminated or damaged due to the nature of the process. FIG. 1 shows an image 10 photographing the surface of a material on which characters are stamped in a dot type.

Embodiments of the present invention can recognize (20) a character from the image 10 as shown in FIG. Hereinafter, embodiments of the present invention will be described. However, the embodiments of the present invention are not limited only to recognizing a recognized character as shown in FIG. 1 or recognizing a dot-type character, and embodiments of the present invention are included in a method for recognizing a character. It should be understood based on the configuration of the device that recognizes each action or character.

Referring to FIG. 2, a character recognition method according to an embodiment of the present invention will be described. Although a flow chart for a case in which an input image has already been generated is described in FIG. 2, the present embodiment may further include the step of generating the input image by photographing a subject in which characters are printed or characters are stamped. . The present embodiment may further include displaying the photographed image and receiving designation of a character display area on the displayed image from the user. When the character display area is designated by the user, the possibility of occurrence of an error in character recognition can be reduced by recognizing the character only in the character display area.

A feature value is extracted from the input image (S100). The extracted feature value may be used as basic data to determine a font of a character in the input image (S102). Before extracting the feature value, a pre-process may be performed on the input image so that the feature value can be well extracted from the input image. For example, the pre-processing may include level adjusting. For example, you can brighten colors and adjust contrast.

In this embodiment, various elements used as feature values of an image may be extracted. For example, an edge or contour of the input image may be extracted. In addition, a histogram of a color value of each pixel constituting the input image may be calculated.

The edge of the extracted input image may be used to determine the size of each character included in the input image. In general, since the outline of a character composes a remarkable color change compared to the surroundings, when a predetermined edge extraction logic is applied to an image including the character, the outline of the character forms an edge. Therefore, for alphabets excluding numbers (0 to 9) and lowercase letters i and j in which one stroke constitutes one character, an area occupied by one edge is considered to be a character area. (Also, characters composed of strokes including dot-shaped characters can be separated according to the quality of the input image, so if the edge spacing is within the standard value, it is regarded as the same character area.)

In the present specification, an edge refers to a set of pixels having an uninterrupted reference number and size or more, which are composed of adjacent pixels.

Position correction can be performed on the extracted edge. For example, edge outline removal and parallelism correction or tilt correction may be performed.

The size of the text can be defined by its width and height.

For example, if the x-axis minimum coordinate of a pixel occupied by one edge is 30, the x-axis maximum coordinate is 100, the y-axis minimum coordinate is 100, and the y-axis maximum coordinate is 200, the character indicated by the connected edge is The size could be determined to be 70 horizontal and 100 vertical.

In the case of Hangul, Hanja, lowercase alphabet i, j, and other characters, two or more strokes constitute one character. Even in this case, the strokes constituting one character will be adjacent to each other. Using this property, an edge group consisting of adjacent edges to a degree that satisfies a predetermined criterion is regarded as constituting one character, and an area occupied by the edge group is considered to be a character area.

For example, the minimum x-axis coordinate of a pixel occupied by the first edge is 40, the maximum x-axis coordinate is 70, the minimum y-axis coordinate is 48, and the maximum y-axis coordinate is 108, and one with the first edge If the x-axis minimum coordinate of the pixel occupied by the second edge of the character is 74, the x-axis maximum coordinate is 84, the y-axis minimum coordinate is 40, and the y-axis maximum coordinate is 115, the first edge and the second edge are The size of the composing character will be determined to be 44 width and 75 length.

Meanwhile, when the number of characters is plural, the size of each character may be determined differently. In this case, the average value of each character size can be used.

The extracted edge of the input image may also be used to determine the thickness of a stroke constituting each character included in the input image. Since the edge can be viewed as the outline of each stroke, the thickness of the stroke can be obtained by analyzing the coordinates of each pixel constituting the edge. Differences may occur for each stroke for reasons such as an error in the thickness of the stroke. In this case, the average value of the thickness of each stroke can be used.

When the size of the character and the thickness of the stroke are obtained, it is inquired whether there is any font shape information of each previously stored font that matches the size of the character and the thickness of the stroke. To this end, the present embodiment may further include pre-storing font shape information of each font to be recognized.

The font shape information may include font type, letter spacing, long pitch, character size information, and stroke weight information (in the case of dot type fonts, dot thickness information). The font type indicates whether it is a dot type.

Since the size of the character may vary depending on the distance to capture the character, according to some embodiments, the ratio between the size of the character included in the input image and the thickness of the stroke may be further used as a factor for determining the applied font. For example, if the ratio between the character width of the input image and the thickness of the stroke is 15:1, and the ratio between the character height and the thickness of the stroke is measured as 18:1, it is possible to match this ratio in the font shape information. You can inquire if there is.

In summary, in the present embodiment, extracting a feature value such as an edge from an input image (S100), and determining an applied font for a character displayed in the input image using the feature value and font shape information of each font. Step S102 is performed, and the determining of the applied font includes determining the size of each character displayed in the input image and the thickness of the stroke constituting the character using the extracted feature value, and the It may include the step of determining a font having font shape information corresponding to the size of the character and the thickness of the stroke as the applied font. In this case, the step of determining the size of each character displayed in the input image and the thickness of the stroke constituting the character using the feature value includes an edge having a length less than a predefined minimum length among the extracted edges. The noise processing may include determining a size of each character and a thickness of a stroke using the coordinates of pixels constituting each edge that is not noise processed.

In the present embodiment, the histogram of the color value of the input image may be applied to more accurately extract the edge indicating the outline of the character. As long as each stroke of the character is composed of the same or similar color, and one stroke is composed of the same or similar color, the character color will have a higher frequency in the histogram. By using this point, an edge may be extracted from a character composition candidate pixel having a color value that occupies a ratio of a predetermined value or more in the histogram among pixels constituting the input image and adjacent pixels thereof. In this case, the edge is extracted from a place other than the outline of the character, and the edge is treated as the outline of the character, and as a result, it is possible to prevent the size of the character from being incorrectly determined.

Hereinafter, a method of determining a font according to an embodiment of the present invention when a dot type character is included in an input image will be described.

If the input image contains dot-type characters, the outline of each dot forms an edge. Therefore, when the edge extraction logic is applied to an image including dot type characters, a plurality of circular edges will be extracted. For example, if the edge extraction logic is applied to an image containing a first character composed of 20 dots and a second character composed of 18 dots, 38 circular edges will be extracted. Accordingly, in the present embodiment, when there are more than a predetermined number of circular edges as a result of edge extraction, it is determined that the applied font is one of the dot type fonts.

The shape of the dot is mostly circular, but the shape of the dot may be non-circular, for example, a square or a triangle. In this case, the present embodiment can be generalized by determining the applied font as a dot type font when there are more than a predetermined number of edges of the same shape.

Since the dots constituting one character will be adjacent to each other to the extent that they meet a predetermined criterion, a group is formed from adjacent dots, and the size of the character indicated by the dot group using the area occupied by the formed dot group You will be able to decide.

For example, if the minimum x-axis coordinate of a pixel occupied by the outline edge of each dot in one dot group is 30, the maximum x-axis is 100, the minimum y-axis is 100, and the maximum y-axis is 200 , The size of the character indicated by the dot group may be determined to be 70 horizontally and 100 vertically.

The edge of the extracted input image may also be used to determine the thickness of dots constituting each character included in the input image. Since the edge can be viewed as the outline of each dot, the thickness of the dot can be obtained by analyzing the coordinates of each pixel constituting the edge. A difference may occur for each dot for reasons such as an error in the thickness of the dot. In this case, the average value of the thickness of each dot can be used.

Since the size of the character may vary depending on the distance to capture the character, according to some embodiments, the ratio between the size of the character included in the input image and the thickness of the dot may be further used as a factor for determining the applied font. For example, if the ratio between the character width and the dot thickness of the input image is 22:1, and the ratio between the character height and the stroke thickness is measured as 28:1, whether the font shape information matches this ratio? You can search.

In summary, the steps of determining the applied font include grouping adjacent circular edges, determining the size of each character using the coordinates of pixels constituting each circular edge belonging to a group, and It may include determining the thickness of the dot using the diameter of the circular edge, and determining a dot type font having font shape information corresponding to the size of the character and the thickness of the dot as the applied font.

When a font applied to a character included in an input image is determined through the above-described operation, the input image is classified into each character area image using the font shape information of the applied font (S104). As already mentioned, since the font shape information includes tracking information and long pitch information and size information of each character, the letter spacing and long pitch information can be further reflected in the area information of each character already obtained through edge extraction. A character separation guide line 30 as shown in FIG. 3 may be set in the input image. Thereafter, the input image may be divided into each character area image based on the character separation guide line.

Next, a step of recognizing which character is a character for each character area image is performed (S106). At this time, since the applied font has already been determined, the character area image is compared with each character template of the applied font, and the similarity of each character template of the character area image is calculated, and the character is Determine the character corresponding to the area image. Hereinafter, it will be described in more detail with reference to FIG. 4.

First, image pre-processing is performed on the character area image in order to increase recognition accuracy. A series of image processing, such as increasing sharpness, may be performed, and binarization may be performed (S160). By binarization, the character area image has a pixel value of 0 or 1.

Next, the size of the character area image is matched with the size of the image of each character template of the applied font (S162).

Next, the similarity is calculated by calculating a pixel value matching ratio between the character area image and the image of each character template of the applied font (S164). Since the size of the character area image and the size of the character template image match, pixel values between pixels at the same location may be 1:1 matched. It is assumed that the character template image is also composed of pixels with pixel values of 0 and 1.

Next, the similarity is adjusted according to whether the character area image satisfies the characteristic requirements of each character template of the applied font (S166). Hereinafter, referring to FIG. 5, an operation of comparing the character area image with each character template of the applied font and calculating the similarity of each character template of the character area image will be described in more detail.

5 shows a virtual character area image composed of 5×8 (width×height) pixels and a “0” template of a specific font. According to embodiments of the present invention, as information on a font to be recognized, in addition to the previously described font shape information, template images and feature requirement information of each character may be further stored. 5 shows a template image 36 and feature requirement information 37 for the number '0'. The characteristic requirement information 37 refers to a condition that a pixel value of a specific area of the character area image must be satisfied in order to be recognized as a corresponding character.

First, the similarity of the pixel value of the template image 36 to the character area image 35 and the number '0' is calculated. In this case, it is determined whether pixels at the same location match pixel values, and a ratio of pixels with matching pixel values may be the degree of similarity. Next, an evaluation is performed as to whether the character area image 35 satisfies the feature requirement of the template for the number '0'.

In FIG. 5, three imaginary feature requirements for the number '0' are illustrated. The first feature requirement requires that the pixels in the first area 31 are all zeros. In the case of the character area image 35 to be recognized, this condition is not satisfied. The second feature requirement requires that at least one of the pixels in the second area 32 be 1. In the case of the character area image 35, this condition is satisfied. The third feature requirement requires that at least one of the pixels of the third area 33 is 1. In the case of the character area image 35, this condition is satisfied. That is, in the case of the case shown in FIG. 5, since the character area image 35 did not satisfy only the first characteristic requirement, the numerical value given to the first characteristic requirement is subtracted from the similarity calculated through comparison between images In addition, points may be deducted by the percentage assigned to the first characteristic requirement.

That is, each character template can have a plurality of feature requirements, and each feature requirement can have a different degree of importance. As a result, if the feature requirement is not satisfied, the value or ratio deducted from the similarity is the feature requirement. Each can be assigned differently.

According to an embodiment, the calculated similarity may be increased when the feature requirement is satisfied. That is, when the similarity is adjusted based on whether the character area image satisfies the characteristic requirements of each character template of the applied font, the similarity can be subtracted when the characteristic requirement is not satisfied, or the similarity can be increased when the characteristic requirement is satisfied.

According to another embodiment, as the feature requirement, whether or not a pixel value for a specific region is matched may be used. For example, a character template of the number '0' requires that the pixel value be '0' (e.g., it means white) and the pixel value is '1' (e.g., black The areas 32 and 33 that require color) may be provided as the feature requirements. In this case, the evaluation result of the feature requirement may be calculated based on the degree to which the character region image 35 satisfies the specified pixel values in the regions 31, 32, and 33. In this case, the evaluation result is not determined only in terms of satisfaction and dissatisfaction, but an adjustment value may be determined according to the degree of satisfaction. The adjustment value may have a positive value or a negative value. The adjustment value is reflected in the previously calculated similarity. According to the present embodiment, there is an effect of enabling flexible character recognition even when the character printing or other angle state is atypical.

For one character area image 35, the degree of similarity may be calculated for all characters included in the applied font. As a result, it may be recognized that the character having the highest similarity is the character displayed in the character area image (S168).

Again, the description returns to FIG. 2. The character recognition operation for the character area image described above is repeatedly performed for all character area images included in the input image. As a result, all characters included in the input image can be recognized.

In the character recognition method according to the present embodiment, the applied font is checked according to the feature value of the input image, the input image is separated into a character area image, and each character area image is individually compared with the character template of the applied font, and each character By recognizing characters through the ranking of the similarity with the template, there is an effect of accurately recognizing characters even when the characters are displayed in the input image is bad. In addition, even if a character of a dot type is recognized as a dot type font by accurately recognizing a plurality of dots through edge extraction, the character is accurately identified by comparing the dot type font shape data and image and feature requirements for each character. It has a recognizable effect.

The recognized character or character string may be displayed, stored, or transmitted to an external device through a network interface (S108). The device to which the character recognition method according to the present embodiment is applied may provide a user interface as illustrated in FIG. 6. The user interface may include a display area 40 to check an input image, a text display area 50 to check the recognized result, and a menu area 60 for processing the recognized result. . Through the user interface, the user can collectively process the shooting, recognition, and management of recognition results of material-related information marked on the surface of the material.

A character recognition apparatus according to another embodiment of the present invention will be described with reference to FIG. 7. The character recognition apparatus 70 according to the present embodiment may be a hand-held terminal equipped with a handle to facilitate gripping. The character recognition device 70 includes an image sensing unit 71 that senses an image, a built-in display 72 that displays a user interface as shown in FIG. 6, a control unit (not shown), a storage unit (not shown), and a network. It may include an interface (not shown). The control unit includes a processor for executing a computer program performing the method described with reference to FIGS. 2 to 5 and a storage means for storing an execution code of the computer program.

A character recognition device according to another embodiment of the present invention will be described with reference to FIG. 8. The character recognition apparatus 100 according to the present embodiment may be, for example, a terminal device such as a notebook PC, a desktop PC, a tablet PC, and a smart phone. The character recognition device 100 according to the present embodiment transmits and receives data by being connected to an image acquisition unit 102, a storage unit 104 composed of an auxiliary storage device such as a nonvolatile memory or an HDD, and an external device through a network. It may include a network interface 108, a display 110, and a character recognition unit 106.

The image acquisition unit 102 may include, for example, a lens, an image sensor, and a driving controller. The image acquisition unit 102 converts the optical signal received through the lens into digital data, generates an input image, and provides it to the character recognition unit 106.

The character recognition unit 106 is provided with a processor for executing a computer program that performs the method described with reference to FIGS. 2 to 5 and a storage means for storing the execution code of the computer program, provided from the image acquisition unit 102 Recognizes characters from the input image and generates digital output data indicating the recognized characters. The character recognition unit 106 stores the digital output data in the storage unit 104 to store the digital output data, provides the digital output data to the display 110 to be displayed, or transmits the digital output data to an external device. (108) can be provided.

A character recognition apparatus according to another embodiment of the present invention will be described with reference to FIG. 9. The character recognition apparatus 106 shown in FIG. 9 may be employed as a component in the character recognition apparatus shown in FIG. 8. Accordingly, in FIG. 8, it has been described as being referred to as the character recognition unit 106. 9, the character recognition apparatus 106 according to the present embodiment includes an image feature extraction unit 160, a font storage unit 161, a font determination unit 162, a character region partition unit 163, An image correction unit 164 and a character determination unit 165 may be included.

The image feature extraction unit 160 receives an input image and extracts a feature value. The feature value may be, for example, an edge or a color histogram.

The font storage unit 161 stores font shape information and a template of each character for a font supporting recognition. The character template includes information on character images and character feature requirements. The font shape information may include font type, letter spacing, long pitch, character size information, and stroke weight information (in the case of dot type fonts, dot thickness information). The font type indicates whether it is a dot type.

The font determination unit 162 determines the applied font for the characters displayed in the input image using the feature value and font shape information of each font provided from the font storage unit 161.

The character area division unit 163 divides the input image into each character area image by using the font shape information of the applied font.

The image correction unit 164 performs predetermined image processing for image quality correction on the character area image, performs binarization of the character area image on which image processing has been completed, and converts the binarized character area image to a character determination unit. Provided in (165).

The character determination unit 165 compares the character area image with each character template of the applied font, calculates a similarity for each character template of the character area image, and uses the similarity to calculate a character corresponding to the character area image. Decide.

Each component of FIG. 9 may mean software or hardware such as a field-programmable gate array (FPGA) or an application-specific integrated circuit (ASIC). However, the components are not limited to software or hardware, and may be configured to be in an addressable storage medium or configured to execute one or more processors. The functions provided in the above components may be implemented by more subdivided components, and may be implemented as one component that performs a specific function by combining a plurality of components.

10 is a diagram showing a hardware configuration of a character recognition apparatus according to another embodiment of the present invention. The character recognition apparatus 106 of FIG. 9 may be implemented with a hardware structure as shown in FIG. 10.

Referring to FIG. 10, the data processing device 200 according to the present embodiment includes a system bus 202, a processor 212, a random access memory (RAM) 210, a storage 204, communication with an external device, or It may include an interface unit 206 and a display 208 for connection with the input/output device. The computer program code implementing the character recognition method according to the present invention may be stored in the storage 204, loaded into the RAM 210, and executed by the processor 212. Output data indicating the execution result of the computer program code and the character recognition result is stored in the RAM 210 or the storage 204, transmitted to an external device through the network interface 206, or displayed through the display 208. I can.

The concept of the present invention described with reference to FIGS. 1 to 10 so far may be implemented as computer-readable codes on a computer-readable medium. The computer-readable recording medium is, for example, a removable recording medium (CD, DVD, Blu-ray disk, USB storage device, removable hard disk) or a fixed recording medium (ROM, RAM, computer-equipped hard disk). I can. The computer program recorded on the computer-readable recording medium may be transmitted to another computing device through a network such as the Internet and installed in the other computing device, thereby being used in the other computing device.

Although the embodiments of the present invention have been described with reference to the accompanying drawings, those of ordinary skill in the art to which the present invention pertains can be implemented in other specific forms without changing the technical spirit or essential features. You can understand. Therefore, it should be understood that the embodiments described above are illustrative in all respects and not limiting.

Claims (15)

Extracting a feature value from the input image;
Determining an applied font for a character displayed in the input image using the feature value and font shape information of each font;
Dividing the input image into each character area image by using the font shape information of the applied font;
Comparing the character area image with each character template of the applied font, and calculating a similarity for each character template of the character area image; And
Including the step of determining a character corresponding to the character area image using the similarity,
The step of determining the applied font,
Determining a size of each character displayed in the input image and a thickness of a stroke constituting the character by using the extracted feature value; And
Comprising the step of determining as the applied font a font having font shape information matching the size of the character, the thickness of the stroke, and the ratio between the size of the character and the thickness of the stroke,
Character recognition method. delete The method of claim 1,
Extracting a feature value from the input image,
Including the step of extracting an edge (edge) from the input image,
The step of determining the size of each character displayed in the input image and the thickness of a stroke constituting the character using the feature value,
Noise processing an edge having a length less than a predetermined minimum length among the extracted edges; And
Comprising the step of determining the size of each character and the thickness of the stroke by using the coordinates of the pixels constituting each edge which is not subjected to noise processing
Character recognition method. The method of claim 3,
The step of extracting the edge,
Generating a color histogram of the input image; And
Comprising the step of extracting the edge for a character composition candidate pixel having a color value that occupies a ratio of a predetermined value or more in the histogram among pixels constituting the input image and adjacent pixels thereof,
Character recognition method. The method of claim 1,
Extracting a feature value from the input image,
Including the step of extracting an edge (edge) from the input image,
The step of determining the applied font,
Including the step of determining a dot type font when there are more than a predetermined number of circular edges among the extracted edges,
Character recognition method. The method of claim 5,
The step of determining the applied font,
Grouping adjacent circular edges;
Determining the size of each character using coordinates of pixels constituting each circular edge belonging to one group;
Determining a thickness of a dot using the diameter of each circular edge; And
Comprising the step of determining a dot type font having font shape information corresponding to the size of the character and the thickness of the dot as the applied font,
Character recognition method. The method of claim 1,
The step of calculating the degree of similarity,
Comprising the step of matching the size of the character area image with the size of the image of each character template of the applied font,
Character recognition method. The method of claim 1,
The step of calculating the degree of similarity,
Calculating a similarity by calculating a pixel value matching ratio between the character area image and the image of each character template of the applied font; And
Adjusting the similarity by whether the character area image satisfies the characteristic requirements of each character template of the applied font,
Character recognition method. The method of claim 8,
The character template of the applied font has a first characteristic requirement and a second characteristic requirement,
Adjusting the degree of similarity,
When the character area image does not satisfy the first characteristic requirement, a first value is subtracted from the calculated similarity, and when the second characteristic requirement is not satisfied, a second value different from the first value in the calculated similarity Subtracting the figures,
Character recognition method. The method of claim 8,
The feature requirement is a requirement that a pixel value of a specific area must be satisfied,
Character recognition method. The method of claim 1,
The step of distinguishing,
Setting a character separation guideline in the input image by using the character spacing and long pitch among font characteristic information of the applied font; And
Including the step of dividing the input image into each character area image based on the character separation guide line,
Character recognition method. Determining an applied font for characters displayed in the input image;
Dividing the input image into each character area image by using the font characteristic information of the applied font;
Comparing the character area image with each character template and calculating a similarity of the character area image to each character template; And
Including the step of determining a character corresponding to the character area image using the similarity,
The step of calculating the degree of similarity,
Reflecting whether the character area image satisfies the characteristic requirements of each character template,
The step of determining the applied font,
Determining a size of each character displayed in the input image and a thickness of a stroke constituting the character using the feature values extracted from the input image; And
Comprising the step of determining as the applied font a font having font shape information matching the size of the character, the thickness of the stroke, and the ratio between the size of the character and the thickness of the stroke,
Character recognition method. A computer-readable recording medium in which a computer program for performing the method of claim 1, 3 to 12 is recorded. An image feature extraction unit for extracting a feature value from the input image;
A font determination unit for determining an applied font for the characters displayed in the input image by using the feature value and font shape information of each font;
A character area partitioning unit that divides the input image into each character area image by using the font shape information of the applied font; And
Comprising a character determination unit that compares the character area image with each character template of the applied font, calculates a similarity for each character template of the character area image, and determines a character corresponding to the character area image using the similarity level. But,
The font determination unit,
Using the extracted feature values, the size of each character displayed in the input image and the thickness of the stroke constituting the character are determined,
Determining as the applied font a font having font shape information matching the size of the character, the thickness of the stroke, and the ratio between the size of the character and the thickness of the stroke,
Character recognition device. The method of claim 14,
An image correction unit that performs predetermined image quality correction image processing on the character area image, performs binarization of the character area image on which the image processing is completed, and provides the binarized character area image to the character determination unit More included,
Character recognition device.

Images