JP3541093B2 - Document image inclination detection method and apparatus - Google Patents

Description

[0001]
TECHNICAL FIELD OF THE INVENTION
The present invention relates to a method and an apparatus for detecting the inclination of a document image.
[0002]
[Prior art]
As a conventional method of detecting the inclination of a document image, for example, a method of projecting features in various directions and detecting a peak determined to be most along the line direction (see Japanese Patent Application Laid-Open No. 5-174183). (See Japanese Patent Application Laid-Open No. 2-44486.
[0003]
[Problems to be solved by the invention]
The former method of projecting a feature in a plurality of directions as described above has a problem that a correct angle direction and a peak may be shifted depending on how the feature is taken. In the latter method, it takes time to detect a line in a state where the inclination is unknown.In addition, since different characters are originally compared, comparison cannot be performed at an accurate position. There is a problem that it does not go up.
[0004]
An object of the present invention is to select the same pattern by a method of detecting the same pattern, and compare the positions of the same pattern, thereby comparing the irrelevant character patterns without performing the same. And a method and apparatus for detecting the inclination of a document image.
[0005]
[Means for Solving the Problems]
In order to achieve the above object, in the invention according to claim 1, in a method for detecting the inclination of a document image, a pattern corresponding to a character is cut out from the document image, and a similar pattern corresponding to the same character in the cut-out pattern sequence. Are given the same label, and based on the intersection of a line connecting the pattern of the same label and another pattern, it is determined whether or not the pattern of the same label exists on the same line, and the presence of the pattern on the same line is determined. It is characterized in that the inclination of the image is detected based on the positional relationship between the patterns of the same label.
[0006]
According to the second aspect of the present invention, in the document image inclination detecting apparatus, means for extracting a pattern corresponding to a character from the document image, and assigning the same label to a similar pattern corresponding to the same character in the extracted pattern sequence Means for determining whether or not the pattern of the same label exists on the same line based on the intersection of a line connecting the pattern of the same label with another pattern; Means for detecting the inclination of the image based on the positional relationship between the label patterns .
[0007]
BEST MODE FOR CARRYING OUT THE INVENTION
Hereinafter, an embodiment of the present invention will be specifically described with reference to the drawings.
FIG. 1 shows the configuration of an embodiment of the present invention. In the figure, 101 is an image input unit such as a scanner, 102 is a character pattern cutout unit, 103 is a similar pattern detection unit that extracts a similar pattern from a cutout pattern sequence, and 104 is a similar pattern belonging to the same row. Same row determination means, 105 means for detecting the angle between similar patterns detected in the same row, 106 means angle determination means for determining the angle of the entire image from the angle obtained by each similar pattern, 107 stores various data Is a control unit, and 109 is a data communication path.
[0008]
FIG. 2 is a processing flowchart of the present invention. First, an image is input by the image input unit 101 such as a scanner (step 201). Next, the character pattern cutout means 102 cuts out each character pattern from the input image. This cutout is performed by, for example, integrating adjacent black pixel components. If the tilt angle is unknown, it is difficult to extract characters accurately, especially in Japanese. However, in the present invention, character extraction is not the final object, and even if the character pattern is separated in a partial or side manner, there is no problem in the subsequent processing. Therefore, basically, it is sufficient to extract a connected black pixel component.
[0009]
Next, the similar pattern detection means 103 compares the extracted character patterns, extracts a similar pattern, and performs labeling (step 203). FIG. 3 is a detailed flowchart of step 203. As the pattern matching method in step 303, for example, the test method described in paragraph 9 of US Pat. No. 5,303,313 is used. In other words, this method is a method in which patterns are superimposed while correcting the position, and it is determined whether or not the pixels are similar based on the positions where the pixels appearing different and the patterns of the appearing pixels. According to this test method, when the pattern to be processed matches an existing template, a label attached to the matched template is given to the pattern to be processed (step 304). If not, the pattern to be processed is registered as a new template and a new label is given (step 305). In step 302, in order to reduce the processing time, not all patterns may be processed, and only patterns within a predetermined range in the image may be processed.
[0010]
After the labeling of the similar pattern is completed, the same-line determining unit 104 detects a similar pattern in the same line (Step 204). FIG. 4 is a view for explaining detection of a similar pattern in the same row, and FIG. 5 is a detailed flowchart of step 204.
[0011]
First, a pair extracted as a similar pattern is selected (step 501). In FIG. 4, it is assumed that 401 and 402 are similar patterns provided with the same label. When performing the pattern matching in the processing of the previous step 203, the center position of the pattern is obtained in order to match the position of the pattern overlap. A line segment 404 connects the pattern center positions of the similar patterns 401 and 402. Reference numerals 405 and 406 denote line segments connecting upper left vertices of the circumscribed rectangles of the pattern 401 and pattern 402, and line segments connecting lower left vertices. Then, a rectangle intersecting the line segments 405 and 406 is detected (step 503). These rectangles are 410, 411, and 412 in FIG. Here, since the rectangle 411 is considerably smaller than the current processing target pattern 401, such a rectangle may be excluded from the processing target.
[0012]
Next, of these intersecting rectangles, a rectangle that does not intersect with the line segment 404 connecting the centers of similar patterns is detected (step 504). In the example of FIG. 4, since the rectangles 410, 411, and 412 intersect, there is no corresponding one. At this time, the pair of similar patterns 401 and 402 is determined to be on the same line, and is registered in the list of pattern pairs belonging to the same line (step 505).
[0013]
Returning again to step 501, the next pair of similar patterns is selected. Here, it is assumed that patterns 401 and 403 are similar patterns. A line segment 407 connects the pattern center positions of the similar patterns 401 and 403. 408 and 409 are line segments connecting the upper left and lower left vertices of the circumscribed rectangle of the patterns 401 and 403. As described above, a rectangle that intersects with the line segments 408 and 409 is detected (step 503), and a rectangle that does not intersect with the line segment 407 is detected (step 504). Since the rectangle 413 is detected as the corresponding rectangle, the pair of the pattern 401 and the pattern 403 is not set to the same line (step 506). By repeating the above processing, a pair of patterns existing in the same row can be detected.
[0014]
Returning to FIGS. 1 and 2, the inter-pattern angle detection means 105 detects the detected angle between similar patterns in the same row (step 205). In the example of FIG. 4, the angle is a line segment connecting the centers of the patterns such as the line segment 404. When the angle of each pattern pair is obtained, the entire angle determining means 106 determines the angle of the entire image (step 206). This is the angle of the entire image by taking the average value, median value, etc. of the angles obtained in step 205.
[0015]
In the above-described embodiment, a configuration having dedicated processing units is adopted. However, the present invention is not limited to this, and the units 102 to 106 and 108 are configured by one CPU to perform processing. It is also possible.
[0016]
【The invention's effect】
As described above, according to the present invention , since only the positions between patterns having the same shape are compared, the tilt angle can be detected more accurately. Further, since only the patterns in the same row are compared, the tilt angle can be detected more accurately. Further, since it is possible to determine whether or not the pattern exists on the same line by a simple process, the processing speed is improved.
[Brief description of the drawings]
FIG. 1 is a configuration of an embodiment of the present invention.
FIG. 2 is a processing flowchart of the present invention.
FIG. 3 is a detailed flowchart of step 203 in FIG. 2;
FIG. 4 is a diagram illustrating detection of a similar pattern in the same row.
FIG. 5 is a detailed flowchart of step 204 in FIG. 2;
[Explanation of symbols]
Reference Signs List 101 Image input means 102 Character pattern cutout means 103 Similar pattern detection means 104 Same line determination means 105 Inter-pattern angle detection means 106 Overall angle determination means 107 Data storage unit 108 Control unit 109 Data communication path

Claims (2)

In the document image inclination detection method, a pattern corresponding to a character is cut out from the document image, a similar pattern corresponding to the same character is given the same label in the cut-out pattern series, and a line segment connecting patterns of the same label is provided. And whether the pattern of the same label exists in the same row based on the intersection with another pattern, and determine the inclination of the image based on the positional relationship between the patterns of the same label present in the same row. A document image inclination detecting method characterized by detecting. A document image skew detection device, means for cutting out a pattern corresponding to a character from the document image, means for giving the same label to a similar pattern corresponding to the same character in the cut-out pattern sequence, Means for determining whether or not the pattern of the same label exists on the same line based on the intersection of the line segment connecting with the other pattern, and the positional relationship between the patterns of the same label on the same line. A means for detecting the inclination of the image based on the image data.

Images