JPH06311333A - Picture processing unit - Google Patents

Abstract

PURPOSE:To obtain the picture processing unit in which distortion is corrected in an excellent way over the entire picture and rotation or lateral direction of an input picture or the like is corrected. CONSTITUTION:An area segmentation means 1 segments plural predetermined areas from an input picture in which a distortion correction mark is possibly in existence. A mark detection means 2 detects at least three marks from the picture of the plural areas segmented by the area segmentation means 1. A picture correction means 3 applies correction processing such as tilt or magnification/reduction or the like to the input picture to generate an output picture based on a position relation of the mark detected by the mark detection means 2, information of a kind of a mark or the like and an averaged relation of the marks of plural pairs.

Description

Detailed Description of the Invention

[0001]

BACKGROUND OF THE INVENTION 1. Field of the Invention The present invention relates to an image processing apparatus for correcting the distortion of an image read by, for example, an image scanner or a facsimile by the positional relationship of distortion correction marks arranged in advance in a document image. .

[0002]

2. Description of the Related Art When an image is read by an image scanner or a facsimile, expansion or contraction occurs depending on the state of the original. In addition, misalignment and image tilt may occur depending on the state of the original set. Further, the read image is affected by various causes such as an adjustment deviation of the optical system, a mechanical error of the driving system, or a slip during reading in the document conveying system, and the image is distorted.

Conventionally, as a method of correcting such image distortion, for example, a technique disclosed in Japanese Patent Laid-Open No. 3-15970 is known. In this method, two distortion correction marks are arranged in advance on the left and right sides of the mark sheet, the distortion correction marks are detected, and the inclination of the image is calculated from the positional deviation in the vertical direction to correct the image. .

Another example is Japanese Patent Laid-Open No. 2-378.
There is a method disclosed in Japanese Patent Publication No. 74. In this method, a pair of distortion correction marks are arranged in the main scanning direction in the vicinity of the reading start position of the document, and the distortion correction marks are detected. The image distortion is corrected by scaling the image according to the ratio between the theoretical value of the distance between the two and the distance calculated from the reading resolution.

However, in these conventional techniques, since the correction is performed with the two distortion correction marks, in the case where the distortion is large as in a facsimile machine and the nonlinear distortion exists in the entire image. However, there is a drawback in that even if the correction is properly performed in the vicinity of the marks, the correction is not performed properly in the portion apart from the mark, or in some cases, the distortion is rather increased. Also, when inputting, the original is turned upside down,
In the case of being rotated by a degree, there is a problem that the correction mark itself cannot be detected and the correction process cannot be performed.

[0006]

SUMMARY OF THE INVENTION The present invention has been made in view of the above-mentioned circumstances, and is capable of performing good distortion correction over the entire image, and further transposing the input image, rotating by 90 °, and the like. It is an object of the present invention to provide an image processing device capable of correcting even the above.

[0007]

According to the present invention, in an image processing apparatus, at least a first area extracting means for extracting a plurality of areas in an image and a plurality of areas in the image extracted by the area cutting means are provided. It is characterized by comprising mark detecting means for detecting the mark, the second mark and the third mark, and an image correcting means for correcting an image based on the mark detected by the mark detecting means.

[0008]

According to the present invention, the mark detecting means detects at least the correction marks of the first mark, the second mark, and the third mark, and the detected correction marks are used to form an image in the image correcting means. Is corrected, so that the distortion can be corrected to be smaller than that based on only the relationship between the pair of marks. Further, even if there is a non-linear distortion in the document image, good image correction can be performed.

[0009]

1 is a block diagram showing an embodiment of an image processing apparatus of the present invention. In the figure, 1 is a region cutting means,
Reference numeral 2 is a mark detecting means, and 3 is an image correcting means. The area cutout unit 1 cuts out a plurality of predetermined areas in which the distortion correction marks may exist from the input image. The mark detecting means 2 detects at least three marks from the images of the plurality of areas cut out by the area cutting means 1. The image correction unit 3 performs a correction process on the input image based on the positional relationship of the marks detected by the mark detection unit 2 and the information such as the type of the mark to create an output image.

FIG. 2 is an explanatory diagram showing an example of the positional relationship of the correction marks on the document to be read. On the manuscript,
Correction marks are provided at predetermined positions. In the following description, the case where four correction marks are present on the document will be described as an example. The four correction marks have different shapes, but their vertices (points at which two sides are in contact) are arranged so as to form a rectangular positional relationship. For example, the y coordinates of the vertices of the upper left mark and the upper right mark are arranged at positions having the same value. Also,
The upper left mark and the lower left mark are arranged at positions having the same x coordinate. Each of the four correction marks has, for example, a different shape so that the position on the original document can be recognized.

FIG. 3 is an illustration of an example of four correction marks. All four have different shapes. In the example shown here, all four are hook-shaped correction marks. What is different is the side length of the hook shape, and when the basic length is a,
Each mark is composed of a or a side twice as long as a. For example, the mark placed in the upper left has a length a on both sides, while the mark on the lower left has a horizontal side length a and a vertical side length twice a.

FIG. 4 is an enlarged view of the upper left correction mark. The correction mark is composed of two sides (two lines), but in reality it has a thickness, so the length of the core line is a as shown in the figure, and The length is twice the half of the width, that is, the width is added. Therefore, when detecting the position of the correction mark, it is necessary to set a default value or the like in consideration of such an error due to the line width.

When the correction marks as described above are used, the shapes of the respective correction marks can be detected even when the original is read by being rotated by 90 ° or read upside down. As a result, the orientation of the document can be determined.

The original document as described above is read as an input image by being read by an image scanner or transmitted by a facsimile, for example. The input image may be stored in a storage device such as a disk once, and the image may be read out when processing is performed. Further, the image after other processing such as noise removal may be used as the input image.

FIG. 5 is a flow chart showing an example of the operation of the image processing apparatus of the present invention. When the image of the document with the correction mark as described above is input, in S11, a candidate area in which the correction mark may exist is cut out. Next, in S12, the correction marks are extracted from the cut-out area, the type of each correction mark is determined from the size of the correction mark, and the coordinates of the vertices of the correction mark are obtained. In S13, from the size of the correction mark, that is, the type of the correction mark,
It is determined whether the original image is in the normal direction, for example,
If the image is a landscape image rotated 90 ° or the image is upside down, the image is rotated for correction. In this step, rough image rotation corrections such as 90 ° and 180 ° are performed. In S14, the inclination of the document is obtained from the coordinates of the vertices of the correction mark, and if there is inclination, rotation correction is performed. Further, in S15, the enlargement or reduction ratio of the document is obtained from the coordinates of the vertices of the correction mark,
If enlargement or reduction is necessary, enlargement or reduction correction is performed. The input image is corrected by the processing of S13 to S15.

The individual processes described above will be specifically described below. FIG. 6 is an explanatory diagram of an example of cutting out a region. When an image of a document as shown in FIG. 2 is input, a region considered to include the correction mark is cut out from the input image as a candidate region. This candidate region needs to be a region in which the correction mark can be detected even when the document is laterally oriented, in consideration of the positional deviation caused by the distortion. Alternatively, it is possible to provide a candidate area for the image in a predetermined orientation or an upside down orientation and a candidate area for the image in the horizontal orientation. Further, it is possible to set an area according to the size of the original image.
It is also possible to detect the orientation and size of the original image depending on which area the correction mark exists. As described above, by setting only the area where the correction mark may exist as the detection target of the correction mark, the processing can be performed at a higher speed than searching the correction mark for the entire image.

FIG. 7 is a flow chart for explaining an example of the operation of detecting the correction mark, FIG. 8 is an explanatory view of the moving direction of the pixel of interest, and FIG. 9 is an example of the detection of the correction mark by contour tracing. FIG. Various methods are conceivable for detecting the correction mark from the cut out region. As one method, there is a method of obtaining the position and size of the mark by obtaining the smallest rectangle including the correction mark. FIG. 7 shows an example of the flow of this processing.
The method of tracking the contour of the correction mark is used. This contour tracking method is basically a method of tracing black pixels outside the mark, as in the case of performing chain code encoding. In the process, the minimum value and the maximum value of each of the x coordinate and the y coordinate are updated, and finally the minimum value and the maximum value of the x coordinate and the y coordinate are obtained to obtain the position and size of the mark. There is.

First, in S21, the upper left pixel of the cut out candidate area is set as a target pixel. Then, in S22, it is determined whether or not the pixel is a black pixel. In the case of a white pixel, that pixel is not the mark portion, so S23
At, the pixel of interest is shifted to the right. When the pixel of interest is at the right end, the pixel at the left end one step below is set as the pixel of interest.
If the pixel of interest is a black pixel, it is considered that a part of the correction mark has been detected, and the contour tracking processing of S24 to S27 is performed.

When a black pixel is found, in S24, a variable for detecting the size of the correction mark is initialized. That is, the minimum value and the maximum value of the x coordinate and the y coordinate are set as the x coordinate and the y coordinate of the pixel of interest. In S25, the pixel of interest is moved. At this time, the pixel of interest is moved so that the content of the mark (black pixel) is always seen to the right. Figure 8
In, the pixel with the center circle is the current pixel of interest. The preceding pixel of interest is assumed to be the pixel to which the diagonal line on the left is applied. In this case, the first candidate for the pixel to be the next pixel of interest is the first pixel seen clockwise relative to the immediately preceding pixel of interest, that is, the pixel with the number 1.
If this pixel is a black pixel, this pixel is set as the next target pixel, and the current target pixel is set as the previous target pixel. If it is a white pixel instead of a black pixel, the number 2 further in the clockwise direction,
Check the pixels with the numbers 3 ... If all the pixels of No. 1 to No. 7 are not black pixels, the previous pixel of interest, that is, the pixel of No. 8 is set as a new pixel of interest again.

By carrying out such processing of moving the target pixel, the pixels of the contour of the correction mark are sequentially tracked. For example, in the example shown in FIG. 9, first, when the inside of the cut out region is scanned from the upper left, the black pixel is first found at the point A. Using this point A as a starting point, the black pixels on the outer edge of the correction mark are traced so that the inside of the correction mark, that is, the side with the black pixel is always seen in the right hand. Figure 9
The route indicated by the inner arrow is the tracking route in this case. The correction mark can be detected at the time when the circuit makes one round around the correction mark and returns to the point A again. In S26 of FIG. 7, it is determined whether or not the circuit has made one round around the correction mark. If it has not returned to the original point A, S27
The variable updating in step S25 and the tracking by moving the target pixel in step S25 are repeated.

While the black pixel is being traced, in step S27, the x-coordinate and y-coordinate of the pixel of interest are checked and compared with the minimum and maximum values of the x-coordinate and y-coordinate to determine the respective variables. Update accordingly. For example, from point A,
When the pixel of interest moves to the lower right pixel, the maximum value of the x coordinate and the maximum value of the y coordinate are updated. During black pixel tracking,
By updating these variables, when the tracking of one round is completed, the x-coordinate, as shown in FIG.
The minimum value and the maximum value of the y coordinate are obtained. The width of the rectangle thus obtained, that is, the maximum value of the x-coordinates-the minimum value of the x-coordinates corresponds to the lateral length of the correction mark, and the height of the rectangle, that is, the maximum y-coordinate. The minimum value of the value-y coordinate is the length of the vertical side of the correction mark.

FIG. 10 is an explanatory diagram of an example of detection of transposition of images and the like. From the size of the correction mark detected as described above, it is possible to recognize and correct the case where the image of the document is turned upside down or turned 90 degrees to the side. . As shown in FIG. 3, all four correction marks have different shapes, and none of the two correction marks have the same side length. Therefore, for example, if the upper right area is cut out and the correction mark is extracted, the direction of the original image can be recognized. Theoretically, the length should be 2 a in the horizontal direction and a in the vertical direction as shown in FIG. The correction mark detected from the actual input image is, for example, as shown in FIG.
It can be seen that this is an extraction of the 180 ° -rotated lower left correction mark shown in FIG. 10 (B). As a result, it can be seen that the original is rotated 180 °. Alternatively, it may be configured to detect in which region a certain correction mark exists. The image may be corrected based on this recognition result. A known method can be used as an image rotation method for correction.

By the above processing, the correction mark can be extracted, the image is corrected in the normal direction, and the reference point for correction, that is, the coordinates of the vertex of the correction mark are obtained. Next, the distortion such as the inclination and the expansion / contraction of the image is corrected by the positional relationship between these reference points. FIG. 11 is an explanatory diagram of an example of calculating the inclination. The inclination of the image can be detected by the inclination of the line segment between the vertices. The upper left of the image is the origin,
When the rightward direction is the x-axis and the downward direction is the y-axis, the coordinates of the vertex of the upper left correction mark are (x0, y
0), and the coordinates of the vertex of the upper right correction mark are (x1, y
If it is 1), the inclination is represented by θ in the figure. y0
And y1 are equal and there is no inclination, θ becomes 0.
This θ is calculated by the following equation. θ = arctan ((y0-y1) / (x1-x0))

FIG. 12 is an illustration of an example of a non-linearly distorted image. If the image is tilted uniformly as a whole, the above θ may be simply corrected as the tilt.
However, in a device such as a facsimile machine in which the accuracy of the paper feeding unit is low, the image may have a non-linear inclination (expansion / contraction) that is not uniform. In such a case, as shown in FIG. 12, the inclination differs depending on the place in the image. Therefore, the calculation of the inclination as shown in FIG.
The image may be corrected not only between a pair of two vertices, but between a plurality of pairs of two vertices, and the average value thereof is used as the inclination of the entire image. For example, since there are four vertices, six relationships (angles to be corrected) can be obtained. Let θ be the average value of these. To correct the inclination, the image may be rotated by θ. The method of rotating the image does not matter.

Next, expansion / contraction correction is performed on the result of the inclination correction. The theoretically known distance between the vertices of the correction marks is compared with the actual distance on the image, and the image is enlarged or reduced so that they are equal.
The direction in which enlargement or reduction is possible, that is, the direction in which expansion and contraction can be measured is the direction in which the correction marks are arranged. When there are four correction marks as in the above example, it is possible to perform expansion / contraction correction in the vertical direction, the horizontal direction, and the diagonal direction. However, in a general linear enlargement / reduction algorithm, it is common to perform correction in the vertical direction or the horizontal direction. Also in this case, for example, even when measuring the expansion and contraction in the vertical direction, not only the one side on the left side but also the average value of the two sides including the right side are used to correct the distortion as a whole. Can be reduced. The same applies to the lateral direction. In this way, the input image is corrected for inclination, correction for enlargement or reduction based on the correction mark, and an output image without distortion can be obtained.

In the above example, the hook-shaped mark is used as the correction mark, but the mark may have another shape as long as the reference point can be extracted. Further, as long as the difference in shape between the correction marks can be detected, it is possible to deal with image transposition, 90 ° rotation, and the like.

In the above example, the case where four correction marks are arranged has been described, but the number of marks may be three or more. If there are three correction marks, it is possible to carry out correction in two directions, and it is possible to carry out good image correction as compared with the conventional case of correction in only one direction. It is also possible to provide a larger number of correction marks, and it is possible to perform finer correction. Also,
When there are three correction marks, transposition of the image, detection of 90 ° rotation, and the like can be performed by detecting the position of the area where the correction mark does not exist.

In the above-mentioned example, the linear correction is used for the correction of the enlargement or reduction of the image, but it is also possible to use the non-linear correction method to improve the correction accuracy. However, in this case, the processing amount increases. Further, the inclination correction and the enlargement / reduction correction may not be separately performed, but may be collectively performed by affine transformation or the like. Alternatively, the non-linear conversion may be performed based on the directions of the two sides of the correction mark and the relationship between the correction marks. When there are a large number of correction marks, it is also possible to configure the output image by performing the correction within the partial area, the position correction of the partial area, and the like for each partial area surrounded by three or more marks.

[0029]

As is apparent from the above description, according to the present invention, the correction is performed by the positional relationship of the three or more distortion correction marks arranged in advance in the original, so that a simple pair of marks can be formed. Since the image is corrected not only by the above relationship but also by the average of the relationship between a plurality of pairs of marks, it is possible to correct the expansion and contraction of the paper and the distortion such as an error in feeding the document. Further, there is an effect that a good image can be obtained by solving the problem that the distortion at the point away from the conventional correction mark becomes rather large, and by correcting so that the distortion becomes small as a whole image. . Further, the correction marks can have different shapes, and in this case, the transposition of the image when the image is input can be promptly detected.

[Brief description of drawings]

FIG. 1 is a block diagram showing an embodiment of an image processing apparatus of the present invention.

FIG. 2 is an explanatory diagram of an example of a positional relationship of correction marks on a document to be read.

FIG. 3 is an explanatory diagram of an example of four correction marks.

FIG. 4 is an enlarged view of an upper left correction mark.

FIG. 5 is a flowchart showing an example of the operation of the image processing apparatus of the present invention.

FIG. 6 is an explanatory diagram of an example of cutting out a region.

FIG. 7 is a flowchart illustrating an example of an operation of detecting a correction mark.

FIG. 8 is an explanatory diagram of a moving direction of a pixel of interest.

FIG. 9 is an explanatory diagram of an example of detection of a correction mark by contour tracing.

FIG. 10 is an explanatory diagram of an example of detection of transposition or the like of an image.

FIG. 11 is an explanatory diagram of an example of inclination calculation.

FIG. 12 is an explanatory diagram of an example of a non-linearly distorted image.

[Explanation of symbols]

1 area cutting means, 2 mark detecting means, 3 image correcting means.

Claims (1)

[Claims] 1. An area cutting-out means for taking out a plurality of areas in an image, and at least a first mark and a second mark from the plurality of areas in the image taken out by the area cutting-out means.
Mark detecting means for detecting the second mark and the third mark,
An image processing apparatus comprising image correction means for correcting an image based on the mark detected by the mark detection means.