JP2013061839A - Image processor, image processing method, image forming device and image processing program - Google Patents

Abstract

PROBLEM TO BE SOLVED: To provide an image processor capable of recognizing a ruled line corner with various shapes in image processing of reading a form image to recognize ruled lines.SOLUTION: The image processor capable of reading an image having ruled lines with width and length, and a ruled line corner area formed in a corner area of two ruled lines, comprises: ruled line corner area extracting means for extracting the ruled line corner area from the image; feature quantity extracting means for extracting a feature quantity of an image of the ruled line corner area from the image of the ruled line corner area; feature quantity dictionary for holding a feature quantity corresponding to a plurality of corner patterns of the ruled line corner area; and ruled line corner extracting means for comparing the feature quantity extracted by the feature quantity extracting means with the feature quantity of the corner pattern held in the feature quantity dictionary to calculate a coincidence degree, and extracting the corner pattern with a predetermined coincidence degree as the ruled line corner.

Description

  The present invention relates to an image processing apparatus, an image processing method, an image forming apparatus, and an image processing program.

  In an image processing apparatus such as an MFP (Multi Function Printer) having a scanner function, a form recognition process is known in which a form image including a table is read and processed. In the form recognition process, a ruled line forming a table is extracted from the form image and converted into vector data, or the extracted ruled line is converted into digital data and reused. In order to perform this process, a technique for extracting ruled lines from a form image has already been disclosed.

  According to the technique described in Patent Document 1, ruled lines are extracted using different methods for ruled lines made of solid lines and ruled lines made of non-solid lines such as dotted lines and broken lines. In the case of a solid ruled line, a black pixel connected rectangle is extracted from the form image and is set as a ruled line candidate. Since the ruled line may be divided even in the solid line, an integration allowable parameter is set in advance, and a plurality of black pixel connected rectangles within the range of the integration allowable parameter are integrated and recognized as a solid line ruled line. In the case of non-solid ruled lines, since the pattern is a pattern in which lines having a length specific to the line type are connected at a specific interval according to a specific rule, a histogram of the width, height, and distance between adjacent pixels of the black pixel connection rectangle is taken. And peaks at widths, heights and distances specific to the line type. Therefore, the distance between adjacent black pixel connected rectangles is calculated, the histogram of their width, height and distance is calculated, the black pixel connected rectangles of the width, height and distance showing the peak are integrated, Recognized as a solid ruled line.

  According to the technique described in Patent Document 2, a collapsed binary image and a blurred binary image are generated from the original image, and ruled lines are extracted using these images. Line segment candidate areas are extracted from the collapsed binary image, and the process is performed for each line segment candidate area, referring to the blurred binary image and the original image, and complementing the blurred binary image with the original image. Is called. If the ratio of black pixels in the line candidate area is equal to or greater than a certain value as a result of the complement processing, it is determined that the line candidate area is a solid ruled line. The technique described in Patent Document 2 is based on the assumption that a ruled line made of a solid line is extracted, and does not assume non-solid line ruled lines such as a broken line and a chain line.

  (See Patent Documents 1 and 2)

  However, in the method of Patent Document 1, for example, a histogram of width, height, and distance is created for a non-solid ruled line, and the black pixel connection rectangle is recognized as a ruled line. A corner portion formed by a circle or the like cannot be recognized. This is because the law that shows a peak in a specific width, height, and distance does not hold. In the method of Patent Document 2, a binarized image can be generated for a region including a corner and a corner ruled line such as a rounded corner can be recognized, but a ruled line that is a solid line can be recognized. It is assumed. Therefore, there is a problem that corners having various shapes cannot be recognized, for example, when the intersection of the ruled lines is constituted by rounded corners and the corners are dotted lines.

  The present invention has been made in view of the above points, and in image processing for recognizing a ruled line by reading a form image, an image processing apparatus and an image capable of recognizing a corner of a ruled line having various shapes It is an object to provide a processing method, an image forming apparatus, and an image processing program.

  Therefore, in order to solve the above problems, an image processing apparatus of the present invention is an image processing apparatus that reads an image having a ruled line having a width and a length and a corner ruled line area formed in a corner area of two ruled lines. A corner ruled line region extracting unit for extracting the corner ruled line region from the image; a feature amount extracting unit for extracting a feature amount of the image of the corner ruled line region from the image of the corner ruled line region; A feature dictionary that stores feature quantities corresponding to a plurality of corner patterns in the ruled line region, a feature quantity extracted by the feature quantity extraction unit, and a pattern of the corners that is held in the feature dictionary And a corner ruled line extracting means for calculating the degree of coincidence by comparing with the feature amount and extracting the corner pattern having a predetermined degree of coincidence as the corner ruled line.

  In order to solve the above problem, the present invention may be an image processing method in the image processing apparatus, an image forming apparatus including the image processing apparatus, and an image processing program for causing a computer to execute the image processing method.

  According to the present invention, it is possible to provide an image processing apparatus, an image processing method, and an image processing program capable of recognizing corner portions of ruled lines having various shapes in image processing for reading a form image and recognizing ruled lines. it can.

2 is a hardware configuration diagram of the image processing apparatus 1. FIG. It is an example of a rounded ruled line. It is an example of the ruled line of various shapes. 3 is a functional block diagram of an image processing unit 108. FIG. It is a result display figure of ruled line extraction processing. It is a figure which shows the detail of a ruled line area | region. It is a figure for demonstrating a corner part ruled line area | region. It is an enlarged view of the rounded corner part after a ruled line extraction process. It is a figure which shows a corner part ruled line area | region. It is a figure which shows the matching image with a rounded ruled line in an input image, and the corner | angular part feature-value dictionary. It is an example of a rounded ruled line pattern. It is a flowchart which shows the flow of a process from an input to an output. 3 is a flowchart illustrating a processing flow in an image processing unit. 2 is a hardware configuration diagram of an image forming apparatus 2. FIG.

  Hereinafter, embodiments of the present invention will be described with reference to the drawings. FIG. 1 is a hardware configuration diagram of the image processing apparatus 1. In the following description, components necessary for the description of the present embodiment are shown for convenience.

(Configuration and function of image processing apparatus 1)
The image processing apparatus 1 is an apparatus having a general scanner function. The image processing apparatus 1 includes a CPU 101, a scanner unit 102, an external storage unit 103, an input image memory unit 104, a display unit 105, an operation unit 106, a corner feature quantity dictionary 107, an image processing unit 108, a communication control unit 109, and an output memory. The unit 113 is provided. The external storage unit 103 includes a media unit 111 such as a CD and a DVD, and an HDD 112. The communication control unit 109 is connected to the network 110.

  The CPU 101 controls the image processing apparatus 1. The scanner unit 102 scans a set document. The external storage unit 103 holds an image using the media unit 111 or the HDD 112. The input image memory unit 104 holds an input image acquired from the scanner unit 102, the external storage unit 103, or the communication control unit 109. The display unit 105 displays a preview of the input image held in the input image memory unit 104 or displays a GUI screen, and a user operation is performed on the operation unit 106 on the GUI screen.

  The corner feature quantity dictionary 107 holds in advance the feature quantities of the corners of the ruled lines as a dictionary. The image processing unit 108 extracts ruled lines from the input image held in the input image memory unit 104 and matches the corners of the ruled lines with the feature amounts of the corner feature amount dictionary 107. Details will be described later. The image processing unit 108 generates a binary image from the input image when the input image is a multi-valued image. The output memory unit 113 holds data on ruled lines processed by the image processing unit 108. The communication control unit 109 obtains an input image from the network 110 and distributes processing result information.

(Outline of processing of image processing apparatus 1)
In FIG. 1, an image input from the scanner unit 102, the external storage unit 103, or the communication control unit 109 is held in the input image memory unit 104 and sent to the image processing unit 108. In the image processing unit 108, vertical ruled lines and horizontal ruled lines are extracted, and corner ruled line regions that are corners of the ruled lines are extracted. In the image processing unit 108, the feature amount of the corner ruled line in the corner ruled line region is calculated. Hereinafter, the corner portion when the corner portion of the ruled line is composed of rounded corners is referred to as a rounded ruled line.

  In the corner feature quantity dictionary 107, the feature quantities for the rounded ruled lines of various patterns are calculated in advance and stored as a dictionary for the corners that are the intersections of the vertical ruled lines and the horizontal ruled lines. The feature amount calculated by the image processing unit 108 is compared with the feature amount dictionary held in the corner feature amount dictionary 107. When the corner feature quantity dictionary 107 holds a feature quantity having a predetermined degree of coincidence, the corner pattern corresponding to the feature quantity is recognized as a rounded ruled line in the corner candidate area. The recognized rounded ruled line digital data and the vertical and horizontal ruled line digital data extracted by the image processing unit 108 are accumulated in the output memory unit 113 and reused as necessary.

  With respect to the rounded ruled lines, the feature values of the rounded ruled lines are stored in the corner feature quantity dictionary 107 as a dictionary, and the rounded ruled lines of various shapes are recognized by comparing the rounded ruled lines with the corners of the input image. Can be acquired.

(An example of ruled lines targeted by this embodiment)
A ruled line that is an object of this embodiment will be described. FIG. 2 is an example of a rounded ruled line. A table with rounded corners among tables often found in a form image or the like will be described. In the present embodiment, extraction is performed by dividing into a linear ruled line in such a table and a rounded ruled line at the corner.

  FIG. 3 shows examples of ruled lines having various shapes. In the present embodiment, not only the general rounded ruled line in FIG. 2 but also the rounded part in the broken line ruled line b) can be accurately extracted including all corners. It should be noted that the ruled lines that can be extracted in the present application are not limited to this range, and can be dealt with in ruled lines having a wide variety of corners.

  In addition, when a corner | angular part is comprised by a vertical ruled line and a horizontal ruled line, it is out of the object of a present Example.

(Configuration and function of the image processing unit 108)
Next, functions of the image processing unit 108 will be described. FIG. 4 is a functional block diagram of the image processing unit 108.

  The image processing unit 108 performs processing on the image input in the image processing apparatus 1. The image processing apparatus 108 includes a processing control unit 201, an image input unit 207, and a result output unit 208. The processing control unit 201 includes a binarization processing unit 202, a ruled line extraction processing unit 203, a corner ruled line region extraction unit 204, a feature amount extraction unit 205, and a matching unit 206.

  The image input unit 207 controls reception of an input image from the input image memory unit 104 to the image processing unit 108. The processing control unit 201 performs processing on the input image input from the image input unit 207. The result output unit 208 outputs the digital data of the vertical ruled lines, the horizontal ruled lines, and the rounded ruled lines extracted by the processing control unit 201 to the output memory unit 113.

  When the input image is a multi-valued image, the binarization processing unit 202 acquires a binarized image from the input image. The ruled line extraction processing unit 203 extracts vertical ruled lines and horizontal ruled lines from the input image. The corner ruled line area extraction unit 204 extracts the vertical ruled line area and the horizontal ruled line area from the vertical ruled line and the horizontal ruled line extracted by the ruled line extraction processing unit 203, and is a corner area from these ruled line areas. A corner ruled line region is extracted. The feature amount extraction unit calculates a feature amount by a predetermined method from the input image in the corner ruled line region extracted by the corner ruled line region extraction unit. The matching unit 206 compares the feature amount of the input image in the corner ruled line region extracted by the feature amount extraction unit 205 with the corner feature amount dictionary 107 held in the corner feature amount dictionary 107 to obtain a predetermined value. The feature amount in the corner feature amount dictionary 107 having the same degree of matching is selected.

(Processing of the image processing unit 108)
The ruled line extraction processing unit 203 performs a ruled line extraction process for extracting a vertical ruled line and a horizontal ruled line on the input image input from the image input unit 207. For the ruled line extraction method, the technique of Patent Document 1 already proposed by the present applicant is used. Specifically, for a solid ruled line, a black pixel connected rectangle is extracted from the input image and set as a ruled line candidate. Since the ruled line may be divided even in the solid line, an integration allowable parameter is set in advance, and a plurality of black pixel connected rectangles within the range of the integration allowable parameter are integrated and recognized as a solid line ruled line. In the case of non-solid ruled lines such as dotted lines and chain lines, a histogram between the width, height, and adjacent distance of the black pixel connection rectangle is generated, and the black pixel connection rectangle of the width, height, and distance indicating the peak is integrated. Recognize as a non-solid ruled line.

  When the ruled line extraction processing unit 203 extracts the vertical ruled line and the horizontal ruled line, the corner ruled line region extracting unit 204 obtains the coordinates of the vertical ruled line and the horizontal ruled line, and uses these coordinates as the ruled line region. Extraction of the vertical ruled line and the horizontal ruled line is shown in FIG.

(Rule area)
FIG. 5 is a result display diagram of the ruled line extraction process. Of the rounded ruled lines shown in FIG. 2, the ruled lines including the vertical ruled lines Lm and the horizontal ruled lines Ln excluding the corners are extracted. The extracted vertical ruled lines and horizontal ruled lines are set as ruled line areas. The ruled line area is necessary for obtaining a corner ruled line area described later.

  In the present embodiment, it is assumed that the ruled line area of the vertical ruled line and the ruled line area of the horizontal ruled line have a right-angle positional relationship. When the vertical ruled line and the horizontal ruled line are straight lines and have a right-angled positional relationship, each ruled line region has the same coordinates as the ruled line itself. For example, in the case of the vertical ruled line Lm and the horizontal ruled line Ln in FIG. 5, the ruled line is a straight line and has a right-angled positional relationship, so that each ruled line region is the same as the coordinates of the ruled line itself.

  6A, 6B, 6C, and 6D are diagrams showing details of the ruled line region. When the vertical ruled line Lm2 and the horizontal ruled line Ln2 are not in a right-angle positional relationship as shown in FIG. 6A, the ruled line area is in a right-angled positional relationship based on one side of the ruled line (for example, Lm2 in FIG. 6A). Configure to be As shown in FIG. 6B, in the case of a straight ruled line, a rectangular region including both ends of the ruled line and at a position perpendicular to the reference ruled line is formed. FIG. 6C shows a ruled line area for the ruled line in FIG. The ruled line region Rm2 and the ruled line region Rn2 are configured to have a right-angled positional relationship. Here, the rectangle in FIG. 6B is a circumscribed rectangle.

  Even if the ruled line is not a straight line such as a wavy line, the circumscribed rectangle of the ruled line is extracted as the ruled line region. In this case, the ruled line region is configured to include all of the pixels constituting the ruled line. FIG. 6D is a configuration example of a ruled line region when the ruled line is a wavy line. Here, the circumscribed rectangle of the ruled line is a rectangle circumscribed by the ruled line.

(Corner ruled line area)
A corner ruled line region is extracted from the ruled line region obtained by the ruled line extraction processing unit 203. FIG. 7 is a diagram for explaining a corner ruled line region.

  If the ruled line area of the vertical ruled line and the ruled line area of the horizontal ruled line are not connected, and the interval between the ruled line area of the vertical ruled line and the ruled line area of the horizontal ruled line is within a certain value, a rounded ruled line may exist at the corner. Therefore, the corner ruled line region is extracted by the following procedure.

  In FIG. 5, vertical ruled lines and horizontal ruled lines are extracted, but the corner ruled lines remain without being extracted. Therefore, a corner ruled line region that is a region including a rounded corner is obtained from the ruled line region. FIG. 7 shows a corner ruled line area obtained for the ruled line extracted in FIG. In this embodiment, C1, C2, C3, and C4 are corner ruled line regions.

  A method for obtaining the corner ruled line region will be described. FIG. 8 is an enlarged view of rounded corners after the ruled line extraction process. The rounded corners in the present embodiment are places surrounded by points A and B in the vertical ruled line region Rm and E and F in the horizontal ruled line region Rn.

  In FIG. 9, a corner ruled line region is shown. The corner ruled line region C1 is a rectangle that can be obtained from the coordinates of the points A and E. Round corner ruled lines M at the corners are shown in the corner ruled line region. This rounded ruled line is compared with the corner feature value dictionary 107 held in the corner feature value dictionary 107, and a rounded ruled line that matches the rounded ruled line in the corner feature value dictionary 107 is used in this embodiment. Performs processing that is regarded as a rounded ruled line.

(Overview of matching)
FIG. 10 shows an image diagram of a comparison between the rounded ruled line of the input image in the corner ruled line area and the corner feature quantity dictionary 107. FIG. 10 is a diagram showing a matching image between the rounded ruled line of the input image and the rounded ruled line of the corner feature quantity dictionary 107.

  In FIG. 10A, a corner feature quantity dictionary 107 holds feature quantities of various rounded ruled lines as a corner feature quantity dictionary. FIG. 10A shows an example of pixel data of a rounded ruled line for easy understanding. However, what is actually held in the corner feature quantity dictionary 107 is not pixel data, and values indicating these feature quantities in numerical values. It is.

  FIG. 10B illustrates an example of matching between the corner feature quantity dictionary 107 and a rounded ruled line in the input image. As for the input image, the feature amount is calculated in the same manner as the corner feature amount dictionary 107, the feature amounts are compared with each other in the image shown in the figure, and the rounded ruled line in the corner feature amount dictionary 107 is matched with the input image. Extract as a rounded ruled line. Although matching is shown on the pixel data in this figure, the degree of coincidence is calculated by comparing numerical values that are feature amounts without actually comparing pixel data.

  Returning to FIG. The input image in the corner ruled line area shows a corner that is the intersection of the vertical ruled line and the horizontal ruled line. The feature amount extraction unit 205 calculates the feature amount of the input image in the corner ruled line region. The matching unit 206 compares the feature amount calculated in the feature amount extraction unit 205 with the feature amount dictionary held in the corner feature amount dictionary 107, and the feature amount held in the corner feature amount dictionary 107 is compared. Among them, a feature amount of a rounded ruled line having a predetermined degree of coincidence is selected and recognized as a rounded ruled line in the corner ruled line region. The digital data of the vertical ruled lines and horizontal ruled lines extracted by the ruled line extraction processing unit 203 and the digital data of rounded ruled lines recognized by the matching unit 206 are output to a predetermined output device by the result output unit 208.

  In the image processing unit 108, corner ruled line regions are extracted, compared with a feature value dictionary stored in advance in the corner feature value dictionary 107, and rounded ruled lines of various shapes are recognized by recognizing the rounded ruled lines. Can be recognized about.

(Corner feature value dictionary 107)
Next, the corner feature quantity dictionary 107 will be described. The corner feature quantity dictionary 107 stores corner ruled line specifying data for specifying the corner ruled lines and the feature values of the corner ruled lines in association with each other. The corner ruled line specifying data is, for example, when the ruled line at the corner is an arc when the circle is divided into four, the data specifying the ruled line is the radius r of the circle and the central angle θ (90 degrees in this case). . Further, for example, when the corner ruled line is an elliptical arc when the ellipse is divided into four, the corner ruled line specifying data includes the major axis radius length ra, the minor axis radius length rb, and the central angle. θ (90 degrees in this case). These are examples, and the corner ruled line specifying data includes any data that can specify the ruled line at the corner.

  The corner feature amount dictionary 107 stores corner ruled line specifying data and corner portion feature amounts in association with each other. A method for calculating the corner feature amount will be described later.

(Calculation of feature value)
The corner feature amount dictionary 107 holds feature amounts of various rounded ruled lines as a dictionary. FIG. 11A shows an example of a rounded ruled line pattern. For example, feature quantities are calculated for these rounded ruled lines, and the feature quantities are stored as a dictionary.

  For the calculation of the feature amount, the multilayer direction histogram method disclosed in Patent Document 3 by the applicant is used. Specifically, the corner ruled line region is first divided into 16 parts. Next, among the pixels constituting the ruled line, the pixel adjacent to the white pixel (background pixel) is regarded as a contour pixel of the rounded ruled line, and eight direction codes are attached to the contour pixel. The corner ruled line region is scanned laterally from each side of the frame of the corner ruled line region to the opposite side, and the direction code of the contour pixel appearing next to the white pixel is detected, and the direction code is detected on the scan line. It memorizes what number was detected. Similar processing is performed in the vertical direction.

  For each divided area of the corner ruled line area, a histogram of direction codes for two layers in the horizontal direction and the vertical direction is obtained, and a vector having each histogram value as a component is used as a feature quantity of the rounded ruled line. In the case of this embodiment, the corner ruled line region is divided into 16 and the direction codes in 8 directions are divided into 2 layers to calculate the histogram, so that 256-dimensional (= 16 × 8 × 2) feature vectors as a whole are obtained. Extracted. The number of divisions, the number of directions, and the number of layers are arbitrary.

  A feature amount is calculated for each rounded ruled line pattern held as a dictionary, and the feature amount is held in the corner feature amount dictionary 107 as a dictionary. For the input image, the feature amount is calculated by the same method for the rounded ruled line included in the corner ruled line region.

(Feature matching)
Matching between the rounded ruled line of the input image and the corner feature quantity dictionary 107 is performed by comparing the numerical values of the feature quantities calculated from the respective corners. Matching is performed using the technique of the above-mentioned patent document 3 and the patent document 5 filed by the applicant. Each feature amount is defined as a feature vector, and a distance between the feature vectors is calculated. A degree of coincidence of the distance with respect to a feature vector to be compared in the corner feature amount dictionary 107 is a certain value (for example, 90%) or more. If so, both ruled lines are considered to match. A rounded ruled line in the corner feature value dictionary 107 corresponding to the matched feature value is regarded as a rounded ruled line of the input image and output.

  Regarding the matching, the technique disclosed in Patent Document 6, which is an application by the present applicant, can also be used. That is, the corner feature dictionary 107 is clustered with feature vectors, the feature vector of the rounded ruled line of the input image is matched with the representative of each cluster, and the cluster with the representative that shows the highest degree of matching is the effective cluster. And Next, the distance between each feature vector in the effective cluster and the feature vector of the rounded ruled line of the input image is calculated, and the feature amount indicating the highest matching degree and having a matching degree equal to or greater than a certain value is included in the input image. It is regarded as a rounded ruled line corresponding to the rounded ruled line and output. When the feature amount held in the corner feature amount dictionary 107 is enormous, clustering in advance can reduce the load on the computer during matching.

  By setting the data held in the corner feature value dictionary 107 as feature values that are numerical values instead of pixel data, the data amount can be reduced. Further, by adopting a method for calculating the degree of coincidence from the feature amount for matching, matching processing can be performed with less load than when comparing pixels.

(Flow from input to output)
Next, the flow of processing from input to output will be described. FIG. 12 is a flowchart showing the flow of processing from input to output.

  In S 101, image data is input to the input image memory unit 104. In addition to acquiring an input image by scanning a document, image data is acquired by various methods such as data captured by a digital camera or the like, or image data acquired via a network. The input image is sent to the image processing unit 108 and received by the processing control unit 201 via the image input unit 207.

  If the input image is a multi-valued image, the binarization processing unit 202 generates a binarized image in S102. The generation of the binarized image is performed using a general technique in addition to the technique described in Patent Document 7 by the present applicant. The generated binarized image is subjected to the above-described processing in the ruled line extraction processing unit 203, the corner ruled line region extraction unit 204, the feature amount extraction unit 205, and the matching unit 206 in S103, and matches the rounded ruled line of the input image. A rounded ruled line in the corner feature quantity dictionary 107 is selected. The rounded ruled lines are regarded as the rounded ruled lines of the input image, and the digital data of each ruled line and the digital data of the rounded ruled lines are output from the result output unit 208 in S104. Each output data is accumulated in the output memory unit 113 and used as necessary.

(Processing flow in the image processing unit 108)
A flow of processing in the image processing unit 108 will be described. FIG. 13 is a flowchart showing the flow of processing in the image processing unit 108. In this description, it is assumed that the input image is a binary image.

  In S201, the ruled line extraction processing unit 203 extracts vertical ruled lines and horizontal ruled lines from the input image. In S202, the ruled line area of the vertical ruled line and the horizontal ruled line is extracted, and the coordinate value is obtained. Based on this coordinate value, the corner ruled line area is extracted by the corner ruled line area extracting unit 204 in S203. As a result of extracting the ruled line area in S202, if the ruled lines are connected or if the interval between the extracted vertical ruled lines and the horizontal ruled lines exceeds a certain value, the process is performed assuming that there is no rounded ruled line at the corner. finish.

  In step S204, the feature amount extraction unit 205 calculates the feature amount for the rounded ruled line in the input image in the corner ruled line region by the method described above. In S205, matching is performed between the feature quantity of the rounded ruled line in the obtained input image and the feature quantity in the corner feature quantity dictionary 107 held in the corner feature quantity dictionary 107, and the matching unit 206 performs matching. Is calculated. A rounded ruled line in the corner feature quantity dictionary 107 having a matching degree exceeding 90%, for example, is regarded as a rounded ruled line in the input image, and is output from the result output unit 208 in S206.

  In calculating the feature amount, data normalization is performed as necessary.

  A variety of rounded ruled line patterns are held in advance as numerical values that are feature quantities, feature quantities are calculated for the corners of the input image, the degree of coincidence is calculated, and the matching feature quantities are rounded ruled lines of the input image I reckon. With this configuration, it is possible to extract various rounded ruled lines with a minimum amount of data for extracting ruled lines having rounded ruled lines.

(Modification)
In this embodiment, it is assumed that a rounded ruled line whose corners are rounded is extracted. However, even in the case where the corner is a straight line as shown in FIG. However, as shown in FIG. 11C, when the vertical ruled line and the horizontal ruled line intersect at a right angle, the ruled line regions are connected to each other and are not assumed in this embodiment.

  In the present application, an image forming apparatus including the image processing apparatus having the above-described function may be used. FIG. 14 shows a hardware configuration diagram of the image forming apparatus 2.

  The image forming apparatus 2 is a general multifunction machine having a scanner function. The image forming apparatus 2 includes functions included in the image processing apparatus 1 and includes an ARDF (Auto Reverse Document Feeder) 301, an image conversion unit 302, a FAXmodem 303, a mailer 304, a memory 305, a USB 306, a printer engine 307, and a power supply unit 308. Prepare.

  In the present embodiment, when a document is included in the document image read by the ARDF 301, a ruled line is extracted through the operation of the operation unit 106, and a feature amount is calculated for the rounded ruled line included in the corner ruled line region. The corner feature quantity dictionary 107 stored in the memory 305 or the HDD 112 is compared with the feature value of the rounded corner pattern, and the rounded ruled line of the input image is extracted. The extracted digital data of the ruled lines and the digital data of the rounded ruled lines are accumulated in the memory 305, transmitted to the outside by the communication control unit 109, or mailed by the mailer 304, and a predetermined output corresponding to the operation is performed. Is called. When the extracted ruled line is formed on paper, it is printed by the printer engine 307 and output.

  As mentioned above, although this invention has been demonstrated based on each embodiment, this invention is not limited to the requirements shown in the said embodiment. With respect to these points, the gist of the present invention can be changed without departing from the scope of the present invention, and can be appropriately determined according to the application form.

DESCRIPTION OF SYMBOLS 1 Image processing apparatus 2 Image forming apparatus 107 Corner part feature-value dictionary 108 Image processing part 202 Binarization processing part 203 Ruled line extraction process part 204 Corner part ruled line area extraction part 205 Feature quantity extraction part 206 Matching part 307 Printer engine

Japanese Patent Laid-Open No. 04-291478 JP 2002-133426 A JP-A 63-778 JP-A-10-105648 Japanese Patent Laid-Open No. 62-22187 JP 11-96304 A JP 2000-331118 A

Claims (10)

In an image processing apparatus that reads an image having a ruled line having a width and a length and a corner ruled line area formed in a corner area of two ruled lines,
A corner ruled line region extracting means for extracting the corner ruled line region from the image;
Feature amount extraction means for extracting the feature amount of the image of the corner ruled line region from the image of the corner ruled line region;
A feature dictionary that stores feature values corresponding to a plurality of corner patterns of the corner ruled line region;
The corner portion having a predetermined degree of coincidence is calculated by comparing the feature amount extracted by the feature amount extracting means with the feature amount of the pattern of the corner portion held in the feature amount dictionary. Corner ruled line extracting means for extracting the pattern of the above as the corner ruled line;
An image processing apparatus comprising: In the case where the read image is a multi-value image, it has a binary image generation means for generating a binary image from the read multi-value image,
The image processing apparatus according to claim 1, wherein the feature amount extraction unit extracts the feature amount from the binary image.   The corner ruled line area extracting means extracts the corner ruled line area based on a vertical ruled line area having a vertical ruled line that is a vertical ruled line and a horizontal ruled line area having a horizontal ruled line that is a horizontal ruled line. The image processing apparatus according to claim 1 or 2. The vertical ruled line area is a circumscribed rectangular area of the vertical ruled line, and the horizontal ruled line area is a circumscribed rectangular area of the horizontal ruled line,
The image processing apparatus according to claim 3, wherein the vertical ruled line region or the horizontal ruled line region is configured to be at a right angle with respect to either the vertical ruled line or the horizontal ruled line. In an image processing method for reading an image having a ruled line having a width and a length and a corner ruled line area formed in a corner area of two ruled lines,
A corner ruled line region extraction procedure for extracting the corner ruled line region from the image;
A feature amount extraction procedure for extracting a feature amount of the image of the corner ruled line region from the image of the corner ruled line region;
The feature amount of the corner pattern held in the feature amount dictionary that holds the feature amount extracted by the feature amount extraction procedure and the feature amount corresponding to the plurality of corner pattern of the corner ruled line region. The corner ruled line extraction procedure for calculating the degree of coincidence and extracting the corner pattern having a predetermined degree of matching as the corner ruled line;
An image processing method comprising: When the read image is a multi-valued image, a binary image generation procedure for generating a binary image from the read multi-valued image is included.
6. The image processing method according to claim 5, wherein the feature amount extraction procedure extracts the feature amount from the binary image.   7. The image processing method according to claim 5, wherein the corner ruled line region extraction procedure is performed based on a vertical ruled line region having a vertical ruled line and a horizontal ruled line region having a horizontal ruled line.   8. The image processing method according to claim 7, wherein the vertical ruled line region and the horizontal ruled line region are configured to include both end points of the ruled line and the ruled line when the vertical ruled line or the horizontal ruled line is an oblique line. An image processing apparatus according to any one of claims 1 to 4,
A printer engine for forming an image on paper based on image data subjected to image processing by the image processing apparatus;
An image forming apparatus comprising:   An image processing program for causing a computer to execute the image processing method according to any one of claims 5 to 8.

Images