KR100667156B1 - Apparatus and method for character recognition by selecting character region in camera document image captured by portable camera - Google Patents

Abstract

1. TECHNICAL FIELD OF THE INVENTION

The present invention relates to a character recognition apparatus and method through character region selection of a character image obtained by a portable camera.

2. The technical problem to be solved by the invention

SUMMARY OF THE INVENTION An object of the present invention is to provide a character recognition apparatus and method for improving character extraction and recognition performance by selecting a character region from a character image obtained by a portable camera and recognizing a character after local binarization.

3. Summary of Solution to Invention

According to an aspect of the present invention, there is provided a character recognition apparatus through character area selection of a character image obtained by a portable camera, comprising: text image input means for receiving a text image obtained by using a portable camera; Image preprocessing means for preprocessing the camera text image received through the text image input means to improve the image; Edge detection means for detecting an edge image with respect to the character image improved by the image preprocessing means; Pixel dilation means for dilating the pixel with respect to the edge image detected by the edge detection means; A virtual character region extraction means for extracting a virtual character region forming a long line having a height above a predetermined threshold by labeling the pixel region with respect to the character image expanded by the pixel expansion means; Binarization means for performing binarization on the character region extracted by the virtual character region extraction means; Character extraction means for extracting a character from the character area binarized by said binarization means; Feature extracting means for extracting a feature of a character extracted by said character extracting means; And character recognition means for recognizing the character using the feature of the character extracted by the feature extraction means.

4. Important uses of the invention

The present invention is used in the character recognition technology field of the pattern recognition field of the image processing technology field.

Portable Camera, Character Image Acquisition, Character Area Selection, Local Binarization, Character Recognition

Description

Apparatus and method for character recognition by selecting character region in camera document image captured by portable camera}             

1 is a block diagram of a character recognition device through the selection of the character region of the character image obtained by the portable camera according to an embodiment of the present invention,

2 is a flowchart illustrating a character recognition method through selecting a character region of a character image acquired by a portable camera according to an embodiment of the present invention;

3A and 3B are diagrams illustrating a text image obtained by a portable camera.

4 is a diagram illustrating a text image preprocessed according to the present invention;

5A to 5C illustrate an edge image, a pixel expansion image, and an image in which only a text area part is selected according to the present invention;

6A and 6B are diagrams illustrating comparison between binarization results of selected text areas in a camera text image and binarization results of an entire camera text image.

* Explanation of symbols for the main parts of the drawings

100: character recognition device 110: character image input unit

120: image preprocessor 130: edge detector

140: pixel expansion unit 150: virtual character area extraction unit

160: local binarization unit 170: character extraction unit

180: feature extraction unit 190: character recognition unit

The present invention relates to a character recognition apparatus and a method for selecting a text area of a text image obtained by a portable camera, and more particularly, to a mobile device such as a PDA (Personal Digital Assistants) or a mobile phone. A character recognition apparatus and method for recognizing a character by performing local binarization on a selected character region after automatically selecting a character region in a character image.

Until now, character recognition has been mostly used by scanning paper documents with scanners and then converting them into text files or electronic documents. However, due to the recent development of camera technology, it is possible to utilize camera functions in mobile devices such as PDAs and mobile phones to obtain information. In addition, as the use of mobile devices increases, the demand for text image processing acquired by a camera increases for the convenience of general users.

Camera character recognition has the advantage of easily acquiring any type of character information existing in the field, unlike a scanner which has mainly inputted paper documents. In other words, camera character recognition is distinguished from conventional scanner-based character recognition in that it can easily acquire and recognize not only paper documents but also characters that cannot be input by the scanner, and then use the recognition results. Camera character recognition targets include tourist information, various explanatory texts, monuments, signs, business cards, and menus. The camera character recognition is one of the character recognition fields that are being actively researched recently because there is an advantage of easily acquiring and recognizing characters with a portable camera regardless of the medium in which the input target characters are recorded.

However, the camera text image, unlike the scanner text image, has a disadvantage of being affected by the ambient light because the condition for inputting the text image is not limited. Therefore, it is known that the camera text image is more difficult to recognize than the conventional scanner-based text recognition. Conventionally published methods for camera character recognition are as follows.

In the first conventional method, a character written on a street sign is acquired by a PDA camera and then recognized. In this conventional method, the user directly selects only the text area in the background of the acquired image, and then transmits the text image to the server. The resulting character recognition results were used to provide information back to the user by translating the information into foreign languages.

The second conventional method relates to the database of paper documents using a camera. Here, the document image is extracted after the character is extracted through binarization. However, the second conventional method does not specifically propose a processing method for preprocessing and text extraction for camera document image recognition.

On the other hand, in general, the camera character image recognition has a problem that it is not possible to guarantee a stable recognition performance for the camera character recognition due to the variety of character input and the character type to be recognized in an unrestricted environment.

Therefore, there is a demand for a method of properly extracting and recognizing a text area from a camera text image to be robust to ambient lighting effects. In addition, memory or processing time has been pointed out as a problem in order to perform a camera character recognition function in a mobile device. Therefore, there is a need for a camera character recognition technology that can compensate for the above problems to ensure a stable character recognition rate.

The present invention has been proposed to solve the above problems and to meet the above requirements, by selecting a character region from a character image acquired by a portable camera and recognizing the character after local binarization, character extraction and recognition performance can be improved. It is an object of the present invention to provide a character recognition device and a method thereof.

That is, the present invention selects a text area from a text image obtained by using a portable camera attached to a mobile device, etc., finds out the virtual text size, determines the sub window size to be binarized from this information, and then recognizes the text after local binarization. It is an object of the present invention to provide a character recognition apparatus and method capable of improving character extraction and recognition performance, reducing memory, and reducing processing time.

Other objects and advantages of the present invention can be understood by the following description, and will be more clearly understood by the embodiments of the present invention. Also, it will be readily appreciated that the objects and advantages of the present invention may be realized by the means and combinations thereof indicated in the claims.

According to an aspect of the present invention, there is provided a character recognition apparatus through character area selection of a character image obtained by a portable camera, the apparatus comprising: text image input means for receiving a character image obtained by using a portable camera; Image preprocessing means for preprocessing the camera text image received through the text image input means to improve the image; Edge detection means for detecting an edge image with respect to the character image improved by the image preprocessing means; Pixel dilation means for dilating the pixel with respect to the edge image detected by the edge detection means; A virtual character region extraction means for extracting a virtual character region forming a long line having a height above a predetermined threshold by labeling the pixel region with respect to the character image expanded by the pixel expansion means; Binarization means for performing binarization on the character region extracted by the virtual character region extraction means; Character extraction means for extracting a character from the character area binarized by said binarization means; Feature extracting means for extracting a feature of a character extracted by said character extracting means; And character recognition means for recognizing the character using the feature of the character extracted by the feature extraction means.

On the other hand, the method of the present invention, in the character recognition method through the character area selection of the character image obtained by the portable camera, an input step of receiving a character image obtained by using the portable camera; An image preprocessing step of performing image preprocessing so that a character part is correctly extracted with respect to the received character image; Extracting an edge image with respect to the image preprocessed text image; Expanding pixels of the edge image to extract a text area; A text area extraction step of extracting a text area from the text image on which the pixel expansion is performed; A binarization step of performing binarization in the extracted character region; A character extraction step of extracting a character from the binarized character area; Extracting a feature from the extracted text; And a character recognition step of recognizing the extracted character.

As described above, the present invention selects a text area from a text image acquired by a portable camera and applies a local binarization method to the selected text area. For this, the height of the text was obtained from the selected text area and used as the window size for local binarization. In addition, since the present invention applies local binarization to a partial region in which a character is written, it is possible to obtain a better binarization result than a binarization result for an entire character image. In addition, the present invention provides an advantage that as the size of the image to be processed becomes smaller, memory problems and processing time problems accompanying them can be solved simultaneously. Therefore, the present invention can be used in real time recognition because it is mounted on a mobile device such as a mobile phone or a PDA as well as a user computer. That is, the present invention can reduce the memory required for processing the camera text image and shorten the processing time, and thus provides a great advantage that it can be used in a portable device such as a PDA or a mobile phone.

The above objects, features and advantages will become more apparent from the following detailed description taken in conjunction with the accompanying drawings, whereby those skilled in the art may easily implement the technical idea of the present invention. There will be. In addition, in describing the present invention, when it is determined that the detailed description of the known technology related to the present invention may unnecessarily obscure the gist of the present invention, the detailed description thereof will be omitted. Hereinafter, exemplary embodiments of the present invention will be described in detail with reference to the accompanying drawings.

1 is a block diagram of a character recognition apparatus by selecting a character region of a character image obtained by a portable camera according to an embodiment of the present invention.

As shown in FIG. 1, the character recognition apparatus 100 may select a text image input unit for receiving a text image obtained by using a portable camera. 110, an image preprocessor 120 for improving the image by pre-processing the camera character image received through the text image input unit 110. An edge detector 130 for detecting an edge image with respect to the character image improved by the image preprocessor 120, and a pixel expansion for dilating the pixel with respect to the edge image detected by the edge detector 130. A virtual character region extracting unit 150 for extracting a virtual character region having a predetermined width and forming a long line from the character image from the pixel expansion unit 140, the virtual character region extracting unit ( The local binarization unit 160 for performing local binarization on the extracted virtual character region 150 and the words and individual characters using structural feature information of the individual characters in the local binarized character region by the local binarization unit 160. A character extractor 170 for extracting a feature, a feature extractor 180 for extracting a feature of a character extracted by the character extractor 170, and a feature of the character extracted by the feature extractor 180 Used to include the character recognition unit 190 for recognizing the characters after classifying the type of the character.

Specific operations and examples of the respective components will be described in detail with reference to FIGS. 2 to 6.

2 is a flowchart illustrating a text recognition method through text area selection of a text image acquired by a portable camera according to an embodiment of the present invention.

First, a user acquires various types of recognition target characters using a portable camera and transmits them to the character recognition apparatus 100 according to the present invention. Accordingly, the text image input unit 110 of the text recognition device 100 according to the present invention receives the text image (210).

That is, in order to experiment with the recognition performance of various camera text images, the distance between the camera and the text is varied to obtain a text image, and the text image is acquired by a mobile phone camera targeting various text fonts. That is, camera character image samples are obtained by varying the distance between the portable camera and the document in order to obtain various types of recognition target characters. In addition, samples may be input to a character image (document image) in which various character fonts are recorded (see FIG. 3A). Character images used in the simulation are images obtained in an unrestricted lighting condition environment. Therefore, regardless of the medium in which the character is recorded in the indoor and outdoor environment, the character image written on the printed matter is obtained (see FIG. 3B).

However, when inputting a text image obtained by a portable camera, the text part may not be properly recognized due to distortion and focus problems according to the characteristics of the lens. In addition, since a text image obtained by using a low resolution camera is inputted, a text image is often blurred or overlapped with a neighboring character. Therefore, it is very difficult to divide and recognize a text region. Therefore, in the present invention, as described later, the text area is first extracted to proceed with binarization, character extraction, and recognition.

Thereafter, an image preprocessing algorithm is applied to the text portion to properly extract the text portion (220). Because, unlike the text image input to the scanner, the camera text image has a vignette phenomenon and a blurring phenomenon in which the text image is blurred at the edges of the acquired text image due to the influence of ambient light. Since these factors act as factors that cause the misrecognition of characters, a method of extracting and recognizing character parts is required.

Therefore, after converting a color image to a brightness image, image preprocessing is performed by dividing the pixel values located in the partial region in the brightness image evenly over the entire level by using the gray level level normalization method. Emphasizes text areas from the background. As a result, the improved image is shown in FIG. 4.

Thereafter, an edge image is extracted with respect to the image image preprocessed (230). That is, an edge image as shown in FIG. 5A is extracted from the character image to which the image preprocessing algorithm is applied. This is to extract a region in which a character exists in the received character image.

Thereafter, the pixel of the edge image is expanded to extract a text area (240). That is, by connecting discontinuous edge points, the pixels corresponding to the character region are expanded in the binarized image, and the image is shown in FIG. 5B.

Thereafter, a long line region having a predetermined height is extracted as a virtual character region (250). That is, after labeling the pixel area of the character image on which the pixel expansion has been performed, a long line area having a height higher than or equal to a predetermined threshold is regarded as a virtual character area and extracted. The resulting image is shown in FIG. 5C.

Thereafter, local binarization is performed in the extracted virtual character area (260). That is, within the extracted virtual character area, a threshold for binarization is calculated, and local adaptive binarization is performed to extract characters while reducing loss of character stroke as much as possible. As described above, the portable camera character recognition has an advantage of easily acquiring and inputting various kinds of real-world character information with the portable camera, regardless of the medium on which the characters are written, thereby recognizing the corresponding character. However, unlike the character image acquired by the camera, the character image acquired by the camera does not properly recognize the character region due to the influence of ambient lighting. In order to solve this problem, a local adaptive binarization method is applied, which divides a character image into sub-regions and investigates and binarizes the brightness distribution of pixels in each region. In the locally adaptive binarization scheme, the size of the sub-window and the size of the target character for reflecting the information of the neighboring pixels have many associations.

Thereafter, words and individual characters are extracted using a combination and separation algorithm in the localized binarized character region (270). That is, for character segmentation, words are extracted using vertical projection and margin information, and then individual characters are extracted using a combination and separation algorithm. In general, unlike Korean characters, vowels and consonants combine to form letters, so that individual letters can be extracted by applying a combination and separation algorithm on the connected pixels. At this time, the connecting pixels other than the characters, that is, the connecting pixels other than the characters resulting from the table, picture, and local binarization are regarded as noise and removed. In particular, in the case of a text image in which Korean-English is mixed, individual characters are extracted using structural feature information of Korean and English.

Then, the feature is extracted for the extracted individual characters (280). That is, the feature of the individual character is extracted using the network feature, the distance information feature, and the contour information for the individual character recognition.

Thereafter, the extracted individual character is recognized to obtain a recognition result (290). That is, after classifying the type of the extracted individual characters, the characters are recognized. In this case, for the recognition of characters mixed with Korean, English, symbols, and numbers, Korean characters are classified into six types of letters, and English, symbols, and numbers are classified into non-Hangul types.

As described above, the present invention proposed a method of extracting and recognizing a character by acquiring a character image using a portable camera, pre-extracting a virtual character region, and then applying a local binarization suitable for the region. In addition, in the present invention, in order to reduce the influence of the ambient lighting, an image enhancement algorithm is applied and a virtual character region is extracted in advance to selectively apply a local binarization method suitable for the region. In addition, by using the combination and separation algorithm to extract the individual characters to recognize the characters by type, it is possible to reduce the recognition of the error by reducing the burden on the recognizer. In addition, the present invention is to solve the problem that the existing camera character image was considered to be difficult to recognize due to the influence of the ambient light or the camera lens, it is also a problem when using the camera character recognition algorithm in the mobile phone, etc. Processing time and memory issues were solved. In addition, the present invention can improve the camera character recognition performance by performing local binarization by selectively selecting the character region as well as processing time problem, memory problem that the existing camera character recognition should solve. In addition, the present invention can be equipped with a camera character recognition technology in a mobile phone or other mobile devices to enable the actual use of character recognition.

As described above, the method of the present invention may be implemented as a program and stored in a recording medium (CD-ROM, RAM, ROM, floppy disk, hard disk, magneto-optical disk, etc.) in a computer-readable form. Since this process can be easily implemented by those skilled in the art will not be described in more detail.

The present invention described above is capable of various substitutions, modifications, and changes without departing from the technical spirit of the present invention for those skilled in the art to which the present invention pertains. It is not limited by the drawings.

As described above, the present invention selects a text area from a text image obtained by using a portable camera attached to a mobile device, finds a virtual text size, and determines a sub window size to be binarized from this information. Recognition has the effect of improving character extraction and recognition performance, reducing memory and reducing processing time.

That is, in the present invention, in order to recognize the camera character image irrespective of the ambient light and irrespective of the character size, the character string is extracted in advance by finding a region in which the character exists and obtaining the virtual character size first, and then subtracting the character from the virtual character size. By determining the size of the window and performing local binarization on the text area to recognize the text, it is possible to obtain better binarization results than the conventional method, thereby improving the performance of camera character recognition, and also to a portable terminal such as a PDA or a mobile phone. Since the size of the image to be processed is reduced when the camera character recognition is executed in the, the memory required for processing the camera character image can be reduced and the processing time can be shortened.

In addition, since the present invention can reduce the memory required for the camera character image processing as described above, and also shorten the processing time, there is a great advantage that the present invention can be mounted on a portable device such as a PDA or a mobile phone and used in practice.

Claims (13)

delete In the character recognition device by selecting the text area of the text image obtained by the portable camera, Text image input means for receiving a text image acquired using a portable camera; Image preprocessing means for preprocessing the camera text image received through the text image input means to improve the image; Edge detection means for detecting an edge image with respect to the character image improved by the image preprocessing means; Pixel dilation means for dilating the pixel with respect to the edge image detected by the edge detection means; A virtual character region extraction means for extracting a virtual character region forming a long line having a height above a predetermined threshold by labeling the pixel region with respect to the character image expanded by the pixel expansion means; Binarization means for performing binarization on the character region extracted by the virtual character region extraction means; Character extraction means for extracting a character from the character area binarized by said binarization means; Feature extracting means for extracting a feature of a character extracted by said character extracting means; And Character recognition means for recognizing a character by using the feature of the character extracted by said feature extraction means Character recognition device comprising a. The method of claim 2, The image preprocessing means, After converting a color image to a brightness image, image preprocessing is performed by placing pixel values located in a partial region in the brightness image evenly on the entire level by using gray level level normalization. Character recognition device, characterized in that for highlighting the text area from. The method of claim 2 or 3, The binarization means, And a threshold value for binarization in each of the extracted virtual character regions to perform local adaptive binarization to extract characters while reducing the loss of character stroke in the character extracting means as much as possible. The method of claim 4, wherein The character extracting means, Extracting words and individual characters using structural feature information such as network characteristics, distance information characteristics, and contour information according to the Jamo arrangement of individual characters in the localized binarized character region by the binarization means, The character recognition means, Character recognition is characterized by classifying the alphabet into six character types according to the characteristics of the characters extracted by the feature extraction means and classifying each character type by classifying English, symbols, and numbers into non-Hangul type. Device. The method of claim 5, The character extracting means, For character segmentation, the word is extracted using vertical projection and margin information, and then the individual characters are extracted by combining and separating algorithms on the connected pixels. Character recognition device, characterized in that to remove other pixels as regards the noise. In the character recognition method through the selection of the text area of the text image obtained by the portable camera, An input step of receiving a text image obtained by using a portable camera; An image preprocessing step of performing image preprocessing so that a character part is correctly extracted with respect to the received character image; Extracting an edge image with respect to the image preprocessed text image; Expanding pixels of the edge image to extract a text area; A text area extraction step of extracting a text area from the text image on which the pixel expansion is performed; A binarization step of performing binarization in the extracted character region; A character extraction step of extracting a character from the binarized character area; Extracting a feature from the extracted text; And Character recognition step of recognizing the extracted character Character recognition method comprising a. The method of claim 7, wherein The input step, Character recognition method characterized in that for receiving a text image obtained by the mobile phone camera to the character image obtained by the mobile phone camera and various text fonts by varying the distance between the camera and the character. The method of claim 7, wherein The image preprocessing step, After converting a color image to a brightness image, image preprocessing is performed by placing the pixel values located in the partial region in the brightness image evenly on the entire level by using a gray-level normalization method. Character recognition method characterized in that the emphasis from the text area. The method according to any one of claims 7 to 9, The text area extraction step, And labeling the pixel area of the character image on which the pixel expansion has been performed to form a long line having a height higher than or equal to a predetermined threshold value as a virtual character area. The method of claim 10, The binarization step, Character recognition method characterized in that for performing local adaptive binarization by calculating a threshold for binarization in the extracted virtual character area. The method of claim 11, The character extraction step, After the words are extracted using the vertical projection and margin information for character segmentation, the individual characters are classified according to the structural characteristic information such as the network feature, distance information feature, and contour information according to the alphabetic arrangement of the individual characters using a combination and separation algorithm. Character recognition method characterized in that the extraction. The method of claim 12, The character recognition step, Character recognition method characterized in that the character is recognized after classifying the type of the extracted individual characters.

Images