JPH09120430A - Image processor - Google Patents

Abstract

PROBLEM TO BE SOLVED: To speedily correct the tilt of an image by extracting the contour of the image and also the tilt of a rectilinear part such as a horizontal rule, etc., and correcting the tilt of the rectilinear part into a horizontal or vertical state. SOLUTION: The software of a control part 2 extracts the contour of an image stored in a memory 51 and stores it in a memory 52. Then the coordinates of four corners are calculated based on the contour data stored in the memory 52 and stored in a memory 53. The contour data stored in the memory 52 undergo the tilt correction processing and the corrected contour data are stored in a memory 54. An area-filled image is generated from the corrected contour data, and an image undergone its tilt correction is obtained in the memory 51. Then an address contained in a file 7 storing the corrected image is decided by the interaction carried out with a user using a console 6 and stored in the file 7. Thus the contour of an image is extracted and the tilt of a rectilinear part such as a horizontal rule, etc., is extracted. Then the tilt of the rectilinear part is corrected into a horizontal or vertical state.

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, and more particularly to an image processing apparatus suitable for normalizing the inclination of a document image.

[0002]

2. Description of the Related Art In conventional document image processing, particularly binary document image processing, it is usual to perform photoelectric conversion and digitization of an image to convert an input image into a digital image before processing. The device for photoelectric conversion and digitization used at this time is called a scanner. When the human power of the image,
The relative tilt between the document and the scanner is unavoidable, and the resulting tilt of the image adversely affects the document image processing, eg, extraction of character strings in the document. Even if the relative inclination between the document and the scanner can be made extremely small, the character string in the document may be originally inclined with respect to the paper when the document is created by a typewriter or when the document is copied. .

In the prior art, in order to deal with such a problem, the document image is integrated with respect to a plurality of directions, a direction in which the integrated value has a peak value is obtained, and the direction makes an angle with the edge of the paper. A method of rotating a document image is known, and it is reported, for example, in the 1991 Conference of the Institute of Electronics, Information and Communication Engineers of Japan, 1769 "Character cutout method on a form with a black background". However, this method has a problem that the amount of processing becomes enormous because the grayscale values of all the pixels are calculated.

[0004]

SUMMARY OF THE INVENTION An object of the present invention is to provide a document image processing apparatus capable of correcting image inclination at high speed.

[0005]

In order to achieve the above object, the present invention extracts a contour from an image, extracts the inclination of a straight line portion such as a ruled line from the extracted contour, and determines the inclination of the straight line portion. It is characterized in that the inclination of the image is corrected at high speed by correcting the inclination so that it becomes horizontal (or vertical).

FIG. 1 shows the principle. 1A shows an example of a document image, FIG. 1B shows an image obtained by performing contour extraction from FIG. 1A, and FIG. 1C shows an image in which only the outer contour is extracted from FIG. The contour 21 of (B) is obtained corresponding to the ruled line 11 of (A). 21 in (C) shows the same contour. The outline is expressed as a coordinate series when making a round on the black background to the left, and in this case, it can be distinguished as an outer contour if making a round in the counterclockwise direction, and an inner contour if making a round in the clockwise direction.

The contour coordinate series is (Xj, Yj), j = 0,
1, 2, (L-1) (L is the length of the point sequence). At this time, the points giving the maximum value and the minimum value of the two quantities (Xj-Yj) and (Xj + Yj) are obtained to obtain the values shown in FIG.
The four corners of the contour 21 can be obtained as in (D).
However, in (D), the contour 21 is shown enlarged.

By obtaining the four corners of the contour for all the contours and determining the distances between them, only the elongated contour can be extracted. If it is known that there are only (horizontal) ruled lines and text parts on the document, the elongated contour above is a ruled line, and the angle with that horizontal line is also contour 4.
It can be calculated using the coordinates of the corner. By using this angle, the entire document image may be rotated so that the ruled line is horizontal. The ruled line is one example that is easy to understand, but in principle, as shown in FIG. 1C, the contours of each character often have their sides oriented in a specific direction.

The rotation may be a linear rotation of the original image,
The coordinates of the contour coordinate series may be transformed and then returned to the image. Further, when an elongated pattern other than the ruled line exists on the document, only the ruled line can be extracted by examining the linearity of the point sequence connecting the four corners of the contour. In addition, when there are slanted lines and vertical lines as straight lines on the document, similarly obtain the slant of the straight lines and extract only horizontal lines and vertical lines (almost horizontal or nearly vertical lines) and use either of them. The tilt of the image may be corrected at high speed by correcting the tilt.

Further, in the aspect of the present invention, a feature amount is obtained for the contour extracted from the image, the feature amount is integrated in various directions, and a direction in which the integrated value becomes a peak is obtained, The inclination of the image is corrected at high speed by correcting the inclination by the angle formed by the direction with the edge of the paper.

As a specific example of the contour extracted from the image, a rectangle circumscribing the contour (outer contour) is formed,
It can take the length of its vertical or horizontal sides. This method can be used to determine the inclination of a character string when no ruled line exists. As another example, the area of a rectangle or an arbitrary function value obtained based on these values can be considered.

The principle is shown in FIGS. FIG. 2 (A)
Shows an example of a document image, (B) shows an image obtained by performing contour extraction from (A), and (C) shows an image obtained by extracting only the outer contour from (B). A circumscribed rectangle as shown in (D) is created from the outer contour of (C). The coordinates of the circumscribed rectangle can be easily obtained from the maximum value and the minimum value of the outline point sequence coordinates, and these are (PXj, PYj), (QXj, QYj), (RXj,
RYj) and (SXj, SYj). Here, P, Q, R, and S represent the upper left, lower left, lower right, and upper right of the circumscribed rectangle j, respectively. That is, (QXj, QYj) and (RXj, RYj) respectively represent images on the lower side of the circumscribed rectangle j. At this time, the next quantity V (α) is calculated for a certain angle α. First, the array H (k), k = 0, 1, ..., (N-1) that stores the histogram is cleared to zero. This is the vertical length when the N document image is M × N. Next, H (Kj) ← H (Kj) + | QXj−RXj | is calculated by a loop for j, where Kj = G [QYj−QXj · tan (α)]. In the above, the symbol "←" means substitution and G
[X] indicates a function for converting the real number x into an integer.

As shown in FIG. 3, Kj is obtained by drawing a straight line with an angle α from a point Qj at the lower left corner of the 31 circumscribed rectangle j.
It is the Y coordinate of the point that intersects the left side 32 of the image. Therefore, the histogram H (Kj) corresponds to inclining the image by the angle α and integrating the length of the lower side of the circumscribed rectangle j.
This is called a side length peripheral distribution in the angle α direction. When the loop for j is completed, V (α) = Σ | H (k) −H (k−1) | is calculated. V (α) is an amount representing the sharpness of the distribution around the edge length in the direction of the angle α. By changing the angle α in steps of 0.1 ° from −2 ° to + 2 °, and obtaining the maximum angle of V (α), the direction in which the lower sides of the circumscribed rectangles best overlap can be obtained. In general, lines of characters are printed or written with their lower sides aligned, and therefore the angle at which they are arranged on the same line can be considered as the inclination of the character string.

The case where the length of the bottom side of the rectangle is used as the characteristic value will be described with reference to FIG. FIG. 7A is a diagram when the histogram H (K) is obtained at an angle α equal to the tilt angle. The circumscribing rectangle of the character is shown in a square column. As shown in the figure, when the length of the bottom of the rectangle is integrated in the direction equal to the tilt angle, the histogram has a large value at the position corresponding to each character line. If the rectangle indicating the outline of the character is completely regular, the points at the lower left corner of each rectangle are aligned, and the histogram has peaks only at the corresponding positions as shown in the figure. Therefore, V (α) for this is a value obtained by integrating the difference between the values of H (K) and is large, and therefore takes a large value. In reality, the character patterns have different shapes and sizes, but even in that case, the tendency indicated by the histogram does not change.

On the other hand, when the integrating direction α is different from the tilt angle as shown in FIG. 7B, no sharp peak appears and
It has a distribution around the position corresponding to each character line. Therefore, V (α) for this is much smaller than that in FIG. By calculating V (α) in a plurality of directions in this way, the inclination of the image can be known by utilizing the fact that V (α) becomes maximum when α is close to the inclination angle. In this example, the length of the side of the rectangle circumscribing the character is used as the feature amount, but it is clear that the same effect can be obtained even if the area of the rectangle is used as the feature amount.

[0016]

Embodiments of the present invention will be described below in detail with reference to the drawings.

FIG. 4 is a block diagram showing the arrangement of an apparatus adopting the document processing system according to the embodiment of the present invention. Each part of the device is connected to a bus 1, and the overall operation is controlled by a control unit 2. The information (document image) on the document 3 is photoelectrically converted and digitized by the scanner 4 into a digital image, which is stored in the memory 51 via the bus 1. The memory 51 forms a part of the memory 5 together with 52, 53 and 54 described later. Instead of obtaining the digital image 51 from the scanner 4, it may be read from a digital image file device such as an optical disk. Further, in the following description, it is assumed that each pixel is binarized into 1 bit, but even if one pixel is represented by multi-values, or if color information is photoelectrically converted by a color scanner and color information is added, it is possible to perform a small amount of processing. Obviously, it can be applied with only changes.

By the software of the control unit 2, the memory 5
The contour of the image in 1 is extracted and stored in the memory 52. Since a known method may be used for contour extraction, the details are omitted. Instead of the contour extraction, a so-called connected region extraction method may be used. Then, as will be described later, the memory 5
The coordinates of the four corners are obtained from the contour data in 2 and stored in the memory 53. Next, inclination correction processing is performed on the contour data in the memory 52, and the corrected contour data is stored in the memory 54.
To be stored. An image of which the inside is filled is generated from the contour data after the correction, and a tilt-corrected image is obtained in the memory 51. After the correction is completed, the address in the file 7 in which this image should be stored is determined by the dialog with the user using the console 6, and is stored in the file 7.

Details of the four corner extraction and the inclination correction processing will be described with reference to FIG. FIG. 5 is a diagram illustrating the flow of these processes. The processing flow is PAD (Program Analysis D
iagram) format. At 101, the contour data of the document image (in the memory 52) is input. 102 is
It is a loop control regarding the contour i. At 103, the maximum and minimum values of the sum and difference (X (i) -Y (i)) and (X (i) + Y (i)) of the X and Y coordinates are extracted from the contour data of each contour. i is the number on the contour, and the coordinates (PXi, PYi), (QXi, QYi), (RXi,
RYi) and (SXi, SYi) are determined. (Note that it is different from the four corners of the circumscribed rectangle described below).

Reference numerals 104 to 108 are portions for performing ruled line determination processing. At 103, the maximum value R _max is initialized. 10
4 is a loop control for the contour i.

In step 106, the ratio of the width Wi to the height Hi is compared with the maximum value up to that point, and in the case of the new maximum value, R is calculated in 107.
Update _max and register i at that time as im.

At 108, it is determined whether the width Wi and the height Hi of the contour im are within a certain range. When the contour im is within the range, the contour im is a horizontal ruled line (when there are a plurality of lines, it is the longest thin line). Then, the inclination θ of the contour im which is determined to be a ruled line in 109 is determined. However, θ = arctan ((SY
i-RYi) / (SXi-RXi)) Moreover, 110 is an exit at the time of failure.

Reference numerals 111 to 116 are portions for generating an image subjected to inclination correction processing. 111 clears the output image area B. Reference numeral 112 is a loop control for the contour i. Reference numeral 113 is a loop control relating to the point j forming the contour i. Reference numeral 114 is a process for obtaining the coordinates (Xj, Yj) in which the coordinates (Xj, Yj) of the point j are tilt-corrected by the calculation formula Yj ′ = Yj−Xj · tan (θ). 115 is a pixel B (Xj, Y
j) is set to 1, which is a process of mapping the point sequence data of the contour i on the output image area. Reference numeral 116 is a process of filling the inside of the contour, in which the inside is filled with black for the outer contour and the inside H for the inner contour. If the inner contour is filled after filling the contour, the original pattern is completely restored.

As can be understood from the above description, although the tilt correction uses the displacement deformation, the rotation deformation can also be used.

Further, by changing the ruled line determination logic of 108, the upper and lower sides or the left and right sides of the table can be used instead of the ruled lines.

Next, examples of the present invention will be described.

The configuration of the apparatus employing the document processing apparatus according to this embodiment is the same as that of the above-described embodiment, and the control unit 2
The processing in is different. In the present embodiment, inclination correction is performed by detecting the inclination of the character string. The difference from the above-described example is the part of inclination detection, and this part will be described.

FIG. 6 is a flow chart of the inclination detection processing in the second embodiment, and corresponds to the portion for performing the ruled line determination processing of 103 to 107 in FIG. In FIG. 6, 201 to 211 are portions for detecting the inclination of the character string. 201 is V _max
Is set to the initial value. 202 is a loop control relating to the angle α. 203 sets 0 to the initial value of V (α). 204 clears the buffer areas H (k), k = 0.1, ..., (N-1) to zero. Reference numeral 205 is a loop control for the contour i. 206 determines the coordinates (QXi, QYi), (RXi, RYi) of both ends of the lower side of the circumscribed rectangle of the contour i. 207 performs the processing after 208 only when the size of the contour i is within a certain range. 208 is H (Kj) ← H (Kj) + | QXj−RXj | where Kj = QYj−QXj · tan (α) is calculated. When the loop regarding the contour i is completed, the sharpness of the edge length peripheral distribution in the α direction is calculated in 209 by the following formula V (α) = Σ | H (k) −H (k−1) |. 210 and 211 update the value αm of α that gives the maximum value of V (α) and the maximum value V _max . 20
Αm at the end of loop 2 gives the inclination of the character string.

As can be understood from the above description, the ruled lines or the character strings are substantially horizontal in the first and second embodiments. However, even if these are vertical, the processing content is It is easy to make such changes. Also,
Even if it is unknown whether these inclinations are horizontal or vertical, it is easy to select a straight line in the direction close to horizontal or vertical and correct the processing contents so as to correct the inclination.

[0030]

As can be understood from the above description, in the present invention, since the calculation to be performed in the tilt correction is not the whole pixels but only the outline of the black area, the processing amount is significantly reduced. The inclination correction can be performed at high speed.

Further, according to the present invention which pays attention to the character string, it is possible to correct even a document having no ruled line. Further, since the size of the contour is determined and only the contour that looks like a character is integrated and patterns and noise other than the character are excluded from the integration, both high speed and noise resistance can be achieved.

[Brief description of the drawings]

FIG. 1 illustrates the principle of the present invention.

FIG. 2 is a diagram illustrating the principle of the present invention.

FIG. 3 illustrates the principle of the present invention.

FIG. 4 is a block diagram showing the configuration of an apparatus that implements the document processing system of the present invention.

5 is a flow chart of processing in the control unit 2 of FIG.

6 is a flow chart of processing in the control unit 2 of FIG.

FIG. 7 is a schematic diagram illustrating the principle of the present invention.

[Explanation of symbols]

1 ... bus, 2 ... control unit, 3 ... document, 4 ... scanner, 5 ...
Memory, 6 ... Console, 7 ... File.

 ─────────────────────────────────────────────────── ─── Continuation of the front page (72) Inventor Toshihiro Hananoi 2880 Kozu, Odawara-shi, Kanagawa Hitachi Odawara factory (72) Inventor Kiyomichi Kurino 2880 Kozu, Kanagawa prefecture Hitachi Ltd. Odawara factory

Claims (2)

[Claims] 1. Input means for inputting digital image information,
It has a memory for storing the digital image information and a control unit for processing the digital image information in the memory, the control unit extracting the contour data of the digital image information, and the contour data. Image processing apparatus having means for calculating a characteristic amount by calculating the data, means for projecting the characteristic amount in a plurality of directions and integrating to obtain an integrated value, and means for extracting a direction in which the integrated value has an extreme value. . 2. Input means for converting a document into digital image information and inputting it, and a memory for storing the digital image information.
A control unit for processing digital image information in the memory, the control unit calculating means for extracting contour data of character data in the digital image information, and calculating the contour data. An image processing apparatus having means for calculating a characteristic amount, means for projecting the characteristic amount in a plurality of directions and integrating to obtain an integrated value, and means for extracting a direction in which the integrated value has an extreme value.