JP2003187249A - Method for detecting and correcting inclination of image - Google Patents

Abstract

<P>PROBLEM TO BE SOLVED: To exactly detect inclination of an image by enhancing accuracy of feature extraction useful for calculation of an inclination angle of the image. <P>SOLUTION: A plurality of reduced images with different reduction ratio are prepared and the inclination of the images is detected by using information obtained from the reduced images. Characters on documents can be exactly recognized by correcting the inclination of the images based on a detection result by a method for detecting the inclination of the images. The method for detecting the inclination of the images or a method for correcting the inclination of the images are executed conveniently and with satisfactory reproducibility by manufacturing a recording medium in which the method for detecting the inclination of the images or the method for correcting the inclination of the images are recorded, setting the recording medium in a computer system and reading a program recorded in the recording medium. <P>COPYRIGHT: (C)2003,JPO

Description

Detailed Description of the Invention

[0001]

BACKGROUND OF THE INVENTION 1. Field of the Invention The present invention relates to the technical field of document processing, character recognition, filing, and document distribution, and more specifically, it relates to an image skew detection method and apparatus, an image skew correction method and apparatus, and The present invention relates to a recording medium recording these detection methods and a recording medium recording these correction methods. Application fields of the present invention include preprocessing of character recognition, electronic filing, network delivery and the like.

[0002]

2. Description of the Related Art When an image of a document is processed by a character recognition device, the recognition rate is adversely affected if the input image is tilted. Even in an image input device such as a copying machine, a facsimile device, or a scanner, the inclination of the input image is not preferable.

Conventionally, as a character recognition technique, Japanese Patent Laid-Open No. 7-1
Japanese Patent No. 41465 (Title of Invention: Document Image Tilt Detection Method) and Japanese Patent Laid-Open No. 7-105310 (Title of Invention:
Image tilt detection method and table processing method).

The document image inclination detecting method disclosed in Japanese Patent Application Laid-Open No. 7-141465 obtains a circumscribed rectangle of a black pixel connected component from a compressed document image, as claimed in the claims. A rectangle corresponding to a character string is discriminated from the circumscribed rectangle, one or a plurality of reference points are set in each of the discriminated rectangles, and the inclination of the image is determined based on the histogram of the inclination of the straight line connecting the reference points of the adjacent rectangles. In the method for detecting, a histogram of a horizontal proximity rectangle and a histogram of a vertical proximity rectangle are created, a slope and a certainty factor are calculated from each of the created histograms, and the calculated certainty factor and the histogram are used. It is characterized in that either one of the slopes is used based on the obtained information. As described above, in the technique of the above publication, the original image is compressed, the black pixel connected component is extracted from the compressed image, and the character string component is detected to calculate the tilt angle.

On the other hand, the image inclination detecting method described in Japanese Patent Laid-Open No. 7-105310 extracts and extracts a black run having a length equal to or more than a certain value on the input image, as claimed in the claims. The black runs within a predetermined distance of the black runs are integrated into a rectangle that includes all of them, and the inclination of the input image is determined by the mutual positional relationship of the black runs that are integrated when this black run is integrated. It is characterized by determining the direction. As described above, in the technique of the above publication, a black run having a certain length or more is extracted, and neighboring black runs are integrated to extract the ruled line. Then, skew correction (correction of inclination) is performed by obtaining the inclination of the ruled line.

Further, in FIG. 11 of Japanese Patent Laid-Open No. 7-105310, an image correction unit (1002) is provided in a system that detects the direction and angle of inclination of an image.
Corrected binary image memory (1003), frame recognition unit (10
04), the frame coordinate memory (1005), the in-frame character extraction unit (1006), and the character recognition unit (1007) are added to the image tilt correction device. Then, according to the information image inclination correction device, in accordance with the information of the direction and the angle of the inclination output from the inclination detection unit (911), the binary image is formed in the direction opposite to the inclination of the image by the same angle as the inclination angle. By rotating the binary image data in the memory (902), a binary image in which the inclination is corrected is created, and the corrected binary image memory (1
003). In this way
By causing the character recognition unit (1007) to recognize the character image in the corrected binary image memory (1003), the character on the document can be accurately recognized. Note that ()
The numerals inside are the reference numerals described in FIG. 11 of the above publication.

Further, in the techniques disclosed in Japanese Patent Laid-Open No. 7-141465 and Japanese Patent Laid-Open No. 7-105310, a reduced image is created in advance and a predetermined process is performed on the reduced image, thereby reducing the processing amount. Can be reduced.

[0008]

However, in the techniques disclosed in these patent publications, if the reduction ratio of the created reduced image is not appropriate, the accuracy of feature extraction useful for calculating the tilt angle decreases, and There is a problem that the angle is erroneously calculated.

The present invention has been made in view of the above problems of the prior art, and an object thereof is to generate a plurality of reduced images having different reduction ratios to improve the accuracy of feature extraction useful for calculating the tilt angle. This is to realize accurate image tilt detection.

[0010]

The invention according to claim 1 is
A method of detecting a tilt of an image using a reduced image created by using information obtained from the reduced image, wherein a plurality of reduced images having different reduction ratios are used.

According to a second aspect of the invention, there is provided the image inclination detecting method according to the first aspect, wherein the document original image is the object of inclination detection. The invention according to claim 3 is the method for detecting an inclination of an image according to claim 1, wherein the drawing original image is an object of inclination detection. The invention according to claim 4 is characterized in that the information obtained from the reduced image is information regarding a character or a character string.
Alternatively, the image inclination detection method described in 3 is provided. The invention according to claim 5 is characterized in that the information regarding the character or character string includes at least one of the position, size, and arrangement of the character or character string. Is. The invention according to claim 6 is the image inclination detecting method according to claim 1, wherein the information obtained from the reduced image is information regarding a linear component.

According to a seventh aspect of the present invention, there is provided an image inclination correction method characterized in that the image inclination is corrected based on the detection result obtained by the image inclination detection method according to any one of the first to sixth aspects.

An eighth aspect of the present invention is an image inclination detection method, wherein the inclination is detected for the same reading target by using a plurality of images having different reading resolutions.

According to a ninth aspect of the invention, there is provided the image inclination detecting method according to the eighth aspect, wherein the document original image is the object of inclination detection. According to a tenth aspect of the present invention, there is provided the image inclination detection method according to the eighth aspect, wherein the drawing original image is a target of inclination detection. The invention according to claim 11 is the method for detecting an inclination of an image according to claim 8, 9 or 10, characterized in that information about a character or a character string is obtained from a plurality of images having different reading resolutions. The invention according to claim 12 is characterized in that the information regarding the character or character string includes at least one of the position, size, and arrangement of the character or character string. is there. The invention according to claim 13 is
10. Information regarding a linear component is obtained from the plurality of images having different reading resolutions.
0 is a method for detecting the inclination of an image.

The invention according to claim 14 is the invention according to claims 8 to 13.
The image inclination correction method is characterized in that the inclination is corrected based on the detection result of the image inclination detection method described above.

The invention according to claim 15 is the invention as defined in claims 1-6.
An image inclination detecting device, characterized in that the image inclination detecting method according to any one of 8 to 13 is realized.

The invention according to claim 16 is the invention according to claim 7 or 1.
An image tilt correction device is characterized in that the image tilt correction method described in 4 is realized.

The invention described in claim 17 is,
It is a recording medium characterized by recording the image inclination detecting method described in any one of 8 to 13.

The invention according to claim 18 is claim 7 or 1.
4 is a recording medium on which the image inclination correction method described in 4 is recorded.

[0020]

BEST MODE FOR CARRYING OUT THE INVENTION Embodiments of the present invention will be described below with reference to the drawings. First Embodiment The present embodiment relates to claims 1, 8 and 15, and the overall configuration is as shown in the block diagram of FIG. Numerals shown in this figure will be described. 101 is an image input means, 102 is an imaging resolution determination means, 103 is an image reduction rate determination means, 104 is an image reduction means, 105 is a feature extraction means, 106 is an image inclination calculation means, 107. Is a data storage unit, 108 is a control unit for each processing, and 109 is a data transfer path and a control communication path. Further, the flow of the entire processing is shown in FIG. Hereinafter, the description will proceed with reference to FIG.

First, a document image is obtained by the image input means 101 (step S201). The image input means may be a scanner or the like, or may be one that has already been electronically filed. Imaging may be performed in binary, but multi-value in color and gray is also acceptable. If it is not binary, it may be converted into a binary image. A well-known method may be used for conversion from color / gray multi-value to binary.

Next, the image reduction rate determining means 103 determines the reduction rate of the input image (step S203). The image reducing means 104 reduces the input image according to the reduction ratio (step S204). These steps S203, S2
The series of processes of 04 are performed for each reduction rate. Step S
The reduction ratio of 203 is determined by, for example, "1/4 and 1/8".
It may be determined in advance, such as ",", or "1/4, 1/8,
The reduction ratio and the number of reduction ratios may be determined as in “1/16 three types”.

Further, when the image input means 101 can change the resolution and can re-input (input from the scanner, etc.), the following steps S201 and S203 instead of the series of steps S201, S203 and S204. S202
You may perform the series of processes which consist of. That is, the image-capturing resolution determining unit 102 determines the image-capturing resolution of the input image (step S202), then captures an image according to this resolution and inputs the image (step S201). The series of processes in steps S202 and S201 are performed for each resolution. The resolution in step S202 may be determined in advance, for example, "50 dpi and 100 dpi", or only the first one image is read at the designated resolution, and then the predetermined determination is made from the input image. "25 dpi, 50 dpi, 100 dpi
And the number of resolutions may be determined.

In order to make the description consistent, the first one image input at the designated resolution will be referred to as the input image unless otherwise specified, and the image read at the resolution determined in step S202 will be referred to as (input It is called a reduced image (including an image) (a path indicated by a solid line with an arrow on the left side in FIG. 2).

Next (in FIG. 2, the paths shown by the solid lines with arrows on both the left and right sides), the feature extraction means 105 performs feature extraction for detecting the inclination from the reduced image (step S20).
5). The characteristic may be a straight line component such as a character (row) or a ruled line, which is commonly used in the inclination detection processing. Next, the image inclination calculating unit 106 calculates the image inclination (step S206). These steps S2
For 05 and S206, a known method described in JP-A No. 7-105310 or JP-A No. 7-141465 may be used.

The feature extracting means 105 and the image inclination calculating means 106 respectively extract features from a plurality of reduced images of the same original image, and use them to calculate the inclination.

The image reduction ratio / imaging resolution may be determined and the reduced image may be recreated at the timing shown by the broken line arrow in FIG. That is, if it is determined that a feature of another resolution is required after receiving the result of feature extraction or tilt detection, a reduced image may be newly created.

The reduced images obtained by reducing the original image shown in FIG. 3 to 1/4 and 1/16 are shown in FIGS. 4 and 5, respectively. However, since the reduced image is actually small and difficult to see, the size of one pixel is shown enlarged so that each image has the same size. Here, the method of creating the reduced image is described in Japanese Patent Laid-Open No. 7-
According to Japanese Patent No. 141465, the reduction ratio ⅛ is changed, but other methods may be used.

FIG. 3 shows a straight line portion of the document, which is faint and in a broken line state. JP-A-7-
When the straight line is detected by the technique of 105310, it cannot be detected because the straight line is a short broken line in FIG. 4, but it can be extracted in FIG. 5 because the reduction ratio is different from that in FIG.

As described above, according to the present invention, by using a plurality of reduced images, it is possible to extract the features which have been omitted in the prior art. Note that in order to detect a straight line, parameters such as the length of the black run and the integrated distance with the neighboring black run must be set to be significantly different from the case of detecting a general straight line, which lacks versatility. become. It should be noted that by using images captured with different imaging resolutions, it is possible to obtain the same effect as that when using reduced images with different reduction rates.

Next, reduced images obtained by reducing the original image shown in FIG. 6 (this is the character portion of the original) to 1/4 and 1/16 are shown in FIGS. 7 and 8, respectively. However, since the reduced image is actually small and difficult to see, the size of one pixel is shown enlarged so that each image has the same size. Here, the method of creating the reduced image is according to the known technique described in Japanese Patent Laid-Open No. 7-141465 (change the reduction rate ⅛), but other methods may also be used.

When character detection is performed according to Japanese Patent Laid-Open No. 7-141465, a horizontally long component is obtained in FIG. 8 and it cannot be determined whether or not it is a character, but in FIG. It is possible to detect characters. Similar to the case of the straight line component, it is possible to obtain the effect of using an image with a different reduction ratio and imaging resolution.

As described above, according to the conventional technique, even if the feature amount cannot be sufficiently extracted from one reduced image and the tilt angle cannot be calculated, a plurality of reduced images are used as in the present invention. By using the obtained feature amount, it is possible to extract and secure the required feature, so that the tilt angle can be easily calculated.

Second Embodiment The present embodiment (not shown) relates to an image inclination correcting device described in claim 16, and its overall configuration is the same as that of the system shown in FIG. The system configuration has an image correction unit, a corrected binary image memory, a frame recognition unit, a frame coordinate memory, a character extraction unit in a frame, a character recognition unit, and the like.

According to the detection result output from the system (image inclination detection unit) shown in FIG. 1, this image inclination correction device is binary in the direction opposite to the image inclination and at the same angle as the inclination angle. By rotating the binary image data in the image memory, a binary image in which the inclination is corrected is created, and this is stored in the corrected binary image memory. In this way, by causing the character recognition unit to recognize the character image in the corrected binary image memory, it is possible to accurately recognize the character on the document.

Third Embodiment The present embodiment (not shown) relates to the recording medium described in claims 17 and 18, and in the present invention, claims 1 to
It is possible to configure a recording medium on which the image inclination detecting method according to any one of 6, 8 to 13 is recorded, or a recording medium on which the image inclination correcting method according to claim 7 or 14 is recorded.

As an example thereof, there is a CD-ROM in which a program for realizing the processing function of the image inclination detecting method according to the first embodiment is recorded. When the present invention is realized by this CD-ROM, CP
U, ROM, RAM, hard disk, keyboard, C
A general-purpose processing device (computer system) including a D-ROM drive is prepared. And the above CD-R
A desired image inclination can be detected by setting the OM in the CD-ROM drive and performing a predetermined operation by the user. In this case, the data to be processed is document image data input from an image input device such as a scanner.

[0038]

As is apparent from the above description, according to the image inclination detecting method and apparatus of the present invention, it is possible to extract the features for image inclination detection, which cannot be detected by the prior art. Therefore, the accuracy of image tilt detection is significantly improved.

Further, the image tilt correction method according to the present invention
According to the apparatus, for example, by causing the character recognition unit to recognize the character image in the corrected binary image memory, the character on the document can be accurately recognized.

Further, in the present invention, a recording medium on which the above-described image tilt detection method or image tilt correction method is recorded is manufactured,
By setting this in a computer system and reading out the program recorded in this recording medium, it becomes possible to easily and reproducibly carry out the above-described image inclination detection method or image inclination correction method.

[Brief description of drawings]

FIG. 1 is a block diagram showing an overall configuration of a first exemplary embodiment of the present invention.

FIG. 2 is a flowchart showing a processing procedure in the same embodiment.

FIG. 3 is a diagram showing an example of a document image (straight part of document) in the same embodiment.

FIG. 4 is a diagram showing an image obtained by performing straight line detection after reducing the original image of FIG. 3 to 1/4.

FIG. 5 is a diagram showing an image obtained by performing straight line detection after reducing the original image of FIG. 3 to 1/16.

FIG. 6 is a diagram showing another example of a document image (character string of a document) in the same embodiment.

7 is a diagram showing a reduced image obtained by reducing the original image of FIG. 6 to 1/4.

FIG. 8 is a diagram showing a reduced image obtained by reducing the original image of FIG. 6 to 1/16.

[Explanation of symbols]

101 image input means 102 imaging resolution determining means 103 image reduction rate determining means 104 image reduction means 105 Feature Extraction Means 106 image inclination calculating means 107 data storage unit 108 control unit 109 data transfer path and control communication path

Claims (18)

[Claims] 1. A method of detecting a tilt of an image using a reduced image created by using information obtained from the reduced image, wherein a plurality of reduced images having different reduction ratios are used. Detection method. 2. The image inclination detecting method according to claim 1, wherein the document original image is an object of inclination detection. 3. The image inclination detecting method according to claim 1, wherein the drawing original image is an object of inclination detection. 4. The image inclination detecting method according to claim 1, wherein the information obtained from the reduced image is information regarding a character or a character string. 5. The method according to claim 4, wherein the information about the character or the character string includes at least one of the position, size, and arrangement of the character or the character string. 6. The information obtained from the reduced image is information about a linear component.
Alternatively, the image inclination detection method described in 3 above. 7. An image tilt correction method, wherein the image tilt is corrected based on a detection result obtained by the image tilt detection method according to claim 1. 8. A method for detecting an inclination of an image, wherein inclination is detected for the same reading target by using a plurality of images having different reading resolutions. 9. The image inclination detecting method according to claim 8, wherein the document original image is an object of inclination detection. 10. The image inclination detection method according to claim 8, wherein the drawing original image is a target of inclination detection. 11. The image inclination detection method according to claim 8, 9 or 10, wherein information regarding a character or a character string is obtained from a plurality of images having different reading resolutions. 12. The information on the character or character string is
The image inclination detecting method according to claim 11, comprising at least one of the position, size, and arrangement of characters or character strings. 13. The image inclination detecting method according to claim 8, wherein information on a straight line component is obtained from a plurality of images having different reading resolutions. 14. A method for correcting an inclination of an image, which comprises correcting the inclination based on a detection result obtained by the method for detecting the inclination of an image according to claim 8. 15. An image inclination detecting device, characterized in that the image inclination detecting method according to any one of claims 1 to 6 and 8 to 13 is realized. 16. An image tilt correction apparatus, characterized in that the image tilt correction method according to claim 7 or 14 is realized. 17. A recording medium on which the image inclination detecting method according to any one of claims 1 to 6 and 8 to 13 is recorded. 18. A recording medium on which the image tilt correction method according to claim 7 or 14 is recorded.