JP2012128635A - Image processing apparatus, image processing method and program - Google Patents

Abstract

PROBLEM TO BE SOLVED: To determine whether an input image is a captured image or a pure electronic image.SOLUTION: The image processing apparatus includes: image data input means 201 for inputting image data; character rectangle extraction processing means 202 for extracting a character rectangle from the image data; character clustering processing means 203 for clustering a character element from the character rectangle obtained by the character rectangle extraction processing means 202; difference calculation means 204 for superimposing the character element which has been clustered by the character clustering processing means 203 and thereby calculating a difference; and image type determination means 205 for determining whether or not the input image is a captured image indicating the image data which has been optically read depending on whether or not a differential value is a predetermined threshold value or more based on the calculated differential value.

Description

  The present invention relates to an image processing apparatus, an image processing method, and a program for determining whether an input image is a captured image or a pure electronic image.

Conventionally, for example, image data stored in PDF (Portable Document Format) is processed again with OCR (Optical Character Reader), etc., and character recognition data or area identification data is separated. It is used for various purposes, such as being reused for various purposes.
This image data is generally classified into two types: a captured image captured by a scanner, a CCD camera, etc., and image data (hereinafter referred to as “pure electronic image data”) dropped directly into a PDF or image from a document application or image creation application. The processing is not appropriate depending on the type of image data.

In other words, if the image is a captured image, the same processing as a normal image can be performed. However, in the case of a pure electronic image, it does not pass through analog conversion, so that it is noiseless and has no image blur. However, when the conventional recognition system parameter processing is performed on the assumption that noise or the like is included, there is a problem that the image quality is deteriorated or the processing time is increased.
Here, as the parameter processing of the recognition system, for example, processing for correcting image data by skew correction, noise removal, top-and-bottom identification processing, and the like can be cited. However, these processes are not necessary for pure electronic image data, and if they are performed, image quality is deteriorated and useless processing time is required.
Therefore, it is necessary to distinguish and control the captured image and the pure electronic image data. To that end, it is necessary to know whether the processing target image is a captured image.

Regarding this point, an apparatus and a method for knowing whether or not the processing target image is a captured image have been proposed (see Patent Document 1).
That is, Patent Document 1 describes an image processing device that automatically determines whether input image data is a captured image or a pure electronic image. In this image processing device, processing is performed on input image data. Identify the background area as the target area, paying attention to the pixel value of the specified background area, for example, obtain the variance value of the pixel value as to whether the background color RGB is uniform, or there is variation, Depending on whether the variance value is greater than or less than the threshold value, it is determined as a captured image if the variance value is greater than or equal to the threshold value, and is determined to be a pure electronic image if less than the threshold value.

  However, in the image processing apparatus described in Patent Document 1, character rectangles are not extracted as a precondition for determining whether a processing target image is a captured image. In addition to this, a large memory is required.

  The present invention has been made in view of the above-mentioned conventional problems, and its purpose is to determine whether it is a captured image or pure electronic image data from image data in a shorter time than in the past, and An object of the present invention is to provide an image processing apparatus that can reduce the amount of memory.

According to the first aspect of the present invention, there is provided image data input means for performing image data input processing, character rectangle extraction processing means for extracting character rectangles from the image data, and at least obtained by the character rectangle extraction processing means. Character clustering means for clustering character elements from one or more character rectangles, difference calculating means for calculating a difference by superimposing the character elements clustered by the character clustering means, and the difference calculation Based on the difference value calculated by the means, an image type for determining whether the difference value is equal to or greater than a predetermined threshold value is a captured image indicating image data read optically or a pure electronic image And an image processing apparatus.
According to a second aspect of the present invention, in the image processing apparatus according to the first aspect, the image processing for the image data is further performed based on the determination by the image type determination unit. An image processing apparatus having means for switching to processing.
According to a third aspect of the present invention, an image data input step for performing image data input processing, a character rectangle extraction processing step for performing character rectangle extraction processing from the image data, and the character rectangle extraction processing step in the image processing apparatus. A character clustering step of clustering character elements from the obtained at least one or more character rectangles; a difference calculating step of calculating a difference by superimposing the character elements clustered by the character clustering step; Based on the difference value calculated by the difference calculation step, whether the difference value is equal to or greater than a predetermined threshold value determines whether the image is a captured image indicating the optically read image data or a pure electronic image And an image type determination step.
According to a fourth aspect of the present invention, in the image processing method according to the third aspect, the image processing for the image data is further performed for an image for a captured image or a pure electronic image based on the determination in the image type determination step. An image processing method comprising a step of switching to processing.
According to a fifth aspect of the present invention, at least one obtained by an image data input means for performing image data input processing, a character rectangle extraction processing means for performing character rectangle extraction processing from the image data, and the character rectangle extraction processing means. Character clustering processing means for clustering character elements from two or more character rectangles, difference calculation means for calculating a difference by superimposing character elements clustered by the character clustering processing means, and calculation by the difference calculation means Image processing that functions as an image type determination unit that determines whether the image is a captured image or a pure electronic image indicating optically read image data based on whether or not the difference value is equal to or greater than a predetermined threshold value It is a program.
According to a sixth aspect of the present invention, in the image processing program according to the fifth aspect, the computer further performs image processing on the image data based on the determination by the image type determination means for a captured image or a pure electronic image. This is an image processing program that functions as means for switching to image processing.

According to the present invention, when the captured image and the pure electronic image are discriminated from the input image data, the character rectangle is extracted from the input image, and the discrimination processing is performed using the extracted character rectangle. As a result, processing time can be increased and memory can be saved.
In addition, discrimination processing is possible regardless of the resolution of the input image, and even after the discrimination processing, another processing such as character recognition processing can be performed using a character rectangle. Can be reduced.

1 is a configuration diagram of an image processing apparatus according to an embodiment of the present invention. It is a figure which shows the internal structure of the image process part of FIG. It is a flowchart which shows the process sequence of the said image process part. It is a flowchart which shows the subroutine in the flow of FIG. It is a figure explaining the specific example of the process in a difference calculation means and an image kind determination means.

Embodiments of the present invention will be described with reference to the drawings.
FIG. 1 is a configuration diagram of an image processing apparatus according to an embodiment of the present invention.
The image processing apparatus includes a central processing unit (CPU) 101 that performs overall control, a scanner unit 102 that reads a document, an external storage unit 103 such as a hard disk or a CD / DVD / memory card, and an external storage unit 103. An input image memory unit 104 that stores an input image obtained by reading an original file, a display unit 105 that displays a preview of the input image, a panel unit 106 that allows a user to specify an area, and a panel unit 106 that are designated. A region information storage unit 107 that stores the region information; an image processing unit 108 that performs image processing such as character recognition; and a processing result obtained by acquiring an input image from the network 110 or stored in the external storage unit 103 And a communication control unit 109 for delivering the information.

FIG. 2 is a diagram showing an internal configuration of the image processing unit 108 of FIG.
As shown, the image processing unit 108 includes an image data input unit 201, a character rectangle extraction processing unit 202, a character clustering processing unit 203, a difference calculation unit 204, an image type determination unit 205, and an image processing unit. It consists of 206.
Each of the means of the image processing unit 108 is not shown in an internal memory (ROM (Read
It is a function realization means realized by a program stored in Only Memory)). The program is recorded on a known computer-readable recording medium. Further, the computer may be constituted by a CPU 101 and a RAM (Random Access Memory) and ROM (not shown), or may be a computer dedicated to the image processing unit 108.

The image data input unit 201 performs processing for acquiring an image using a scanner, a digital camera, a network, or the like.
The character rectangle extraction processing unit 202 performs processing for extracting a character rectangle from the image data acquired by the image data input unit 201. The character rectangle can be extracted by any method. For example, a circumscribed rectangle of the black run is obtained based on the black run of the generated document image, and a rectangle that can be regarded as a character is extracted from the circumscribed rectangle of the obtained black run. Character rectangles are extracted by performing processing. The determination of a rectangle that seems to be a character element can be executed by limiting the rectangle size based on, for example, the size limit of characters that can be subjected to OCR processing.

The character clustering processing unit 203 performs processing for clustering (separating) the character rectangle data extracted by the character rectangle extraction processing unit 202 into character data of the same character shape as a character element. That is, since the characters have characteristics such as “pluck”, “bias”, and “旁”, the features such as “pluck”, “bias”, and “旁” of the character shape are clustered here.
Here, for a specific method for clustering character rectangles, for example, “a pattern having the same shape and color as a pattern in a reference image is extracted from an image to be processed” described in JP-A-10-208040. Use "method.

  In this method, for example, the identity of the pattern shape and color is determined separately. First, the color difference degree Dc of the two patterns of the character rectangles to be compared is calculated by adding the difference (square) between the input image and the reference image for each RGB over all the pixels in the rectangle, and the calculated color difference When the degree is equal to or less than a predetermined threshold value, it is determined that they are the same, and next, the identity of the shape of the rectangular character pattern is determined by a known pattern matching method (for example, US Pat. No. 5,303,313). For example, the brightness value is calculated from the RGB density values of each pixel of each pattern, the pattern matching process is performed using the brightness data of each pattern, and when it is determined that the pattern matching process is the same, The pattern is determined to be the same pattern.

Also, a known pattern matching process (for example, a pattern matching process described in US Pat. No. 5,303,313) for simultaneously determining the identity of the pattern shape and color is performed for each color of RGB, and all pattern matching is performed. When it is determined that the patterns are the same, a method may be used in which the patterns are determined to have the same shape and color.
Note that the clustering of character images may be performed by the following known method.
JP 2007-058882 “Pattern recognition device”,
JP 2001-338264 “Character recognition pattern dictionary creation device, character recognition pattern dictionary creation method and recording medium”,
JP-A-5-174193 "Character recognition method",
JP-A-4-123283 “Recognition Dictionary Creation Method”.

The difference calculation unit 204 performs processing for calculating the difference by superimposing the data clustered by the character clustering processing unit 203.
Then, the difference is calculated by superimposing the clustered character data. In this case, with respect to the clustered character data, the difference is calculated for all or only part of the clustered character data.
More specifically, as shown in FIG. 5, “A” (“A” (element 1) on the left in the figure is yellow, middle “A” (element 2) is red, and the right “A” (element 3) is colored green). Is referred to a database of features clustered, and differences for all or some of the features are calculated.

The image type determination unit 205 compares the difference value calculated by the difference calculation unit 204 with a predetermined threshold value, and determines that the difference is less than the threshold value, a pure electronic image, and if the difference is greater than or equal to the threshold value, the image type determination image. That is, for example, when the clustered numerical values are within a certain reference value, it is considered that the difference is small and it is determined as a pure electronic image.
In the example of FIG. 5, the determination result of the difference value between another “A” (yellow) of the element 1 of the database and “A” (red) of the clustered element 2 is “low coincidence”, The determination result of the difference value between “A” (green) of element 1 and clustered element 3 is “high degree of coincidence”.

  The image processing unit 206 includes a unit that switches the image processing according to the determination result determined by the image type determination unit 205, thereby performing processing suitable for the determined image. In the image processing of the image processing means 206, in addition to processing suitable for the image such as correction processing of image data such as skew correction, noise removal, and top / bottom identification for the captured image, and processing that omits these processing in a pure electronic image, Processing using character rectangles is also included.

FIG. 3 is a flowchart showing a processing procedure of the image processing unit 108.
In FIG. 3, when image data is input to the main image processing unit 108 (S101), image determination is subsequently performed (S102). After image determination, image processing corresponding to the determination result or character rectangle data is performed based on the determination result. Processing such as character recognition using S is performed (S103).

FIG. 4 is a subroutine of the image determination process in steps S102 and S103 in FIG.
In FIG. 4, when the image data input unit 201 acquires an image by a scanner, digital camera, network, or the like (S201), the character rectangle extraction processing unit 202 extracts a character rectangle from the image data acquired by the image data input unit 201. (S202).
Subsequently, the character clustering processing unit 203 performs clustering on the character data of the same character shape from the character rectangle data extracted by the character rectangle extraction processing unit 202 (S203), and the difference calculation unit 204 performs clustering by the character clustering processing unit 203. The obtained data is superimposed and the difference is calculated (S204).

  The image type determining unit 205 determines the matching degree of the superimposed character rectangle data from the difference value calculated by the difference calculating unit 204 (S205), and if the matching degree is low, that is, the difference value is less than a preset threshold value. For example, it is determined that the image is a captured image, and control is performed assuming the captured image, or a parameter for specifying the captured image is set (S206). Control is performed assuming an image, or a parameter for specifying a pure electronic image is set (S207), and the routine is terminated.

  As described above, in this embodiment, character rectangles are extracted from the input image, and the character rectangles are clustered. For a clustered rectangle, check the degree of matching of the bitmap in the character rectangle. As a method for viewing the degree of coincidence, a clustered character shape is overlapped to look at the difference. Image of processing). These processes can be performed regardless of the resolution of the input image.

  As described above, according to the present embodiment, at the same time as inputting an image, the means (whether by image reading or by a document application) that inputs the image is acquired as a parameter, so that the image is matched. (For example, in the case of a pure electronic image, the character code information is more accurate if it can be extracted from the original electronic information rather than acquiring it from the image information using a technique such as OCR. On the other hand, in the case of a captured image, character code data can be obtained by using OCR).

The character rectangle obtained by this processing can also be used in the next image processing means 206. The next image processing means 206 can perform image processing including processing for identifying areas such as characters, photographs, and tables, character recognition processing, and the like.
In addition, this image processing method can be combined with a technique for recognizing characters for each category of the image data, and can add an application example (such as an effect) that can speed up OCR.

In other words, as already mentioned, if it is a captured image, it may be handled by the same processing as a normal image, but if it is a pure electronic image, analog conversion does not pass, so it is assumed that noise etc. are included When the conventional recognition system parameter processing is performed, there is a problem that the image quality deteriorates and the processing time becomes long.
On the other hand, in the present embodiment, the captured image and the pure electronic image data are controlled separately, that is, the determination is made by looking at the degree of coincidence of the pixels in the character rectangle, and whether or not the captured image is the result is determined. Therefore, for example, pure electronic image data can be processed by correcting image data such as skew correction, noise removal, and top-and-bottom identification. There is an effect of memory.

  DESCRIPTION OF SYMBOLS 101 ... CPU, 102..Scanner unit, 103 ... External storage unit, 104 ... Input image memory unit, 105 ... Display unit, 106 ... Panel unit, 107 ... Storage of area information 108, image processing unit, 109, communication control unit, 110, network, 201, image data input means, 202, character rectangle extraction processing means, 203, character clustering. Processing means 204... Difference calculation means 205... Image type determination means 206.

JP 2008-252862 A

Claims (6)

  Image data input means for performing image data input processing, character rectangle extraction processing means for extracting character rectangles from the image data, and at least one character rectangle obtained by the character rectangle extraction processing means Character clustering processing means for clustering character elements, difference calculation means for calculating a difference by superimposing the character elements clustered by the character clustering processing means, and a difference value calculated by the difference calculation means An image type determination unit that determines whether the difference value is equal to or greater than a predetermined threshold based on whether the image is a captured image indicating the optically read image data or a pure electronic image. Processing equipment. The image processing apparatus according to claim 1,
Further, an image processing apparatus comprising means for switching the image processing for the image data to image processing for a captured image or pure electronic image based on the determination of the image type determination means.   An image data input step for performing image data input processing in the image processing apparatus, a character rectangle extraction processing step for performing character rectangle extraction processing from the image data, and at least one or more obtained by the character rectangle extraction processing step A character clustering step of clustering character elements from the character rectangle, a difference calculating step of calculating a difference by superimposing the character elements clustered by the character clustering step, and calculating by the difference calculating step An image type determination step for determining whether the difference value is equal to or greater than a predetermined threshold based on the determined difference value, whether the image is a captured image or a pure electronic image indicating the optically read image data; An image processing method comprising: The image processing method according to claim 3,
The image processing method further comprises a step of switching the image processing for the image data to image processing for a captured image or a pure electronic image based on the determination in the image type determination step.   An image data input means for performing image data input processing, a character rectangle extraction processing means for performing character rectangle extraction processing from the image data, and a character from at least one character rectangle obtained by the character rectangle extraction processing means. Character clustering processing means for clustering elements, difference calculation means for calculating a difference by superimposing the character elements clustered by the character clustering processing means, based on the difference value calculated by the difference calculation means An image processing program that functions as an image type determination unit that determines whether the image is a captured image or a pure electronic image indicating optically read image data depending on whether or not the threshold value is greater than or equal to a predetermined threshold. In the image processing program according to claim 5,
An image processing program for causing a computer to further function as means for switching image processing for the image data to image processing for a captured image or a pure electronic image based on the determination of the image type determination unit.

Images