JPH0145669B2 - - Google Patents

Description

DETAILED DESCRIPTION OF THE INVENTION (A) Technical Field of the Invention The present invention relates to a contour extraction device for a rectangular object that achieves high processing speed through one-dimensional processing.

(B) Background of the technology One application field of pattern recognition technology is automatic sorting systems for documents, books, etc. In such systems, the surface of the articles is usually Sorting is done based on the recognition results of symbols, characters, patterns, etc. described in the text, but in order to improve the speed and accuracy of the recognition, the outline of the recognition target is recognized in advance as part of the recognition process. The work of extracting the recognition target, so-called "cutting out"
I often do this.

(C) Conventional technology and problems In conventional contour extraction devices for rectangular objects, a two-dimensional calculation filter is used to raster scan a rectangular object and find the maximum density change rate or density change rate at each pixel point. The contour extraction process is performed after determining both the direction in which the density change rate is maximum or the direction in which the density change rate is maximum and the direction in which the density change rate is maximum as a two-dimensional array.

Therefore, the contour extraction process takes a long time and requires a large-capacity register, making it difficult to put it into practical use.

(D) Purpose of the Invention The present invention improves the processing speed of a contour extraction device by converting image information with a two-dimensional structure into a one-dimensional structure at the initial stage of contour extraction processing using a one-dimensional calculation filter. The purpose is to

(E) Structure of the Invention The contour extraction device of the present invention raster scans a digital image of a rectangular object using a one-dimensional calculation filter, and selects a point at which the density change rate is maximum for each scanning line.
an edge detection unit that obtains a point group obtained by extracting points point by point and a point group obtained by extracting points at which the density change rate is minimum one by one as one-dimensional array data; a connecting line generation unit that generates a connecting line by connecting points whose adjacent distances are equal to or less than a predetermined value in the order of scanning lines, and a connecting line obtained by the connecting line generating unit; an edge candidate generation unit that extracts the longest connection line from among the longest connection lines and obtains edge candidates of the rectangular object by linearly approximating the longest connection line; and Among them, an edge element extraction unit extracts connecting lines whose distance from the straight line obtained by the edge candidate generation unit is smaller than a predetermined value, and a linear approximation is performed for all the connecting lines obtained by the edge element extraction unit. a side slope candidate detection unit that detects the slope of the straight line; a slope detection unit that calculates the slope of the rectangular object from the slope candidates for each side of the rectangular object obtained by the side slope candidate detection unit; and the slope detection unit. an edge detection section that detects the sides of the rectangular object from the slope of the rectangular object obtained by the section and all the connecting lines obtained by the side element extraction section; The outline of the rectangular object is extracted based on the slope of the rectangular object obtained after determining the .

In other words, focusing on the fact that the contour to be extracted is made up of a total of four line segments, two of which are orthogonal to each other, a one-dimensional operation filter, or one-dimensional window operation, is performed in two orthogonal directions. The processing is simplified by detecting the edges where the density change rate is maximum or minimum, and the four sets of edges obtained in this way are used as the four sets of contours to be extracted. That is, the inclination of each side is detected by making it correspond to the sides, extracting the parts that become the elements of each side. This processing process requires only a small amount of information to be handled, so the processing is simple and can be performed at high speed. Finally, the slope of the object is determined by weighting the slope of each side and calculating the average.
Together with the previously obtained information regarding the elements of each side, the detection of four sides, that is, the extraction of the contour, can be performed in an extremely short time.

(F) Examples of the invention Hereinafter, the gist of the present invention will be illustrated by way of examples.

The figure shows a system block diagram of an embodiment of the present invention, in which 1 is a noise removal unit that uses a median filter to remove noise components in an input digital image obtained by observing a rectangular object, and 21 is a noise removal unit 1.
The output is raster-scanned in the horizontal direction by a 3×1 differential filter, and the first point at which the density change rate is maximum is extracted one by one for each scanning line.The resulting point group and the density change rate are An edge detection unit extracts the last point that becomes the minimum point one by one and obtains each obtained point group as one-dimensional array data. 22 is an edge detection unit that extracts the last point that becomes the minimum point one by one and obtains each obtained point group as one-dimensional array data. A group of points is obtained by raster scanning and extracting the first point at which the rate of density change is the maximum for each scanning line one by one, and the last point at which the rate of density change is the minimum is extracted one by one. The edge detection units 31 and 32 each obtain the point groups obtained by the edge detection unit 21 as one-dimensional array data, and the adjacent distance is less than or equal to √2 of the scanning line pitch, and a first connecting line generating unit that generates a left side connecting line by connecting points of a point group having five or more constituent points in the order of scanning lines; a second connecting line generating unit that generates a right side connecting line; 33 and 34 are a third connecting line generating unit that generates the upper connecting line from the two sets of points obtained by the edge detecting unit 22 in the same manner as described above, and a fourth connecting line generating unit that generates the lower connecting line. Connecting line generating units 41, 42, 43, and 44 are a first connecting line generating unit 31, a second connecting line generating unit 32, a third connecting line generating unit 33, and a fourth connecting line generating unit 34, respectively.
a first edge candidate generation unit that obtains side candidates of the rectangular object by extracting the longest connecting line from among the connecting lines obtained by and linearly approximating each of the longest connecting lines; The candidate generating section, the third side candidate generating section, and the fourth side candidate generating section, 51, 52, 53, and 54, are the first connecting line generating section 31, the second connecting line generating section 32, and the third connecting line generating section, respectively. Among the connecting lines obtained by the connecting line generating unit 33 and the fourth connecting line generating unit 34, the first side candidate generating unit 41, the second side candidate generating unit 42, and the third side candidate generating unit 43, respectively and a straight line obtained by the fourth edge candidate generation unit 44, a first edge element extraction unit that extracts a connecting line whose distance is smaller than a predetermined value, a second edge element extraction unit, and a third edge. Element extractor and fourth edge element extractor, 61 and 62
, 63 and 64 are the first side element extraction unit 51, respectively.
A third step that linearly approximates all the connecting lines obtained by the second side element extraction section 52, the third side element extraction section 53, and the fourth side element extraction section 54, and detects the slope of the straight line. a first side slope candidate detection unit, a second side slope candidate detection unit, a third side slope candidate detection unit, and a fourth side slope candidate detection unit; 7, a first side slope candidate detection unit 61 and a second side slope candidate detection unit; Each straight line obtained by the side slope candidate detection unit 62, the third side slope candidate detection unit 63, and the fourth side slope candidate detection unit 64 is weighted according to the degree of fitness in each approximation, and The two are tilt detection units that calculate the tilt of the rectangular object from the tilt candidates for each side of the rectangular object by rotating at right angles and calculating the average value; 81, 82, 83, and 84 are respectively the first sides; The point groups constituting the respective connecting lines obtained by the element extraction unit 51, the second edge element extraction unit 52, the third edge element extraction unit 53, and the fourth edge element extraction unit 54 are combined into the former two. For the latter two, scanning is performed using a straight line perpendicular to the slope obtained by the slope detecting section 7, and for the latter two, scanning is performed using a straight line parallel to the slope obtained by the slope detecting section 7, and the distance from the straight line is a first edge for detecting the left, right, upper, and lower contours or edges of the rectangular object, respectively, by searching for the position of the straight line when the number of points smaller than a predetermined value is maximum; They are a detection section, a second side detection section, a third side detection section, and a fourth side detection section.

In this way, the edge detection unit 21 and the edge detection unit 22 convert image information with a two-dimensional structure into a one-dimensional structure, and therefore, all subsequent processing can be performed at high speed by one-dimensional processing. .

(G) Effects of the Invention As explained above, according to the present invention, it is possible to obtain a contour extraction device for a rectangular object with improved processing speed.

[Brief explanation of drawings]

The figure shows a system block diagram of an embodiment of the present invention, in which 21 and 22 are edge detection units;
3 and 34 are first, second, third, and fourth connecting line generation units, respectively; 41, 42, 43, and 44 are first, second, third, and fourth edge candidate generation units, respectively; 51,
52, 53, 54 are the first, second, third, and
Fourth side element extraction unit, 61, 62, 63, 64 are first, second, third, and fourth side slope candidate detection units, respectively; 7 is a slope detection unit, 81, 82, 83, 84
are the first, second, third, and fourth edge detection units, respectively.

Claims (1)

[Claims] 1 Raster scan a digital image of a rectangular object using a one-dimensional calculation filter, and extract one point at a time for each scanning line where the density change rate is the maximum and the point group where the density change rate is the minimum an edge detection unit that extracts points one by one and obtains each point group as one-dimensional array data; and an edge detection unit that extracts points one by one and obtains each point group as one-dimensional array data; a connecting line generation unit that obtains a connecting line by connecting scanning lines in order; and extracting the longest connecting line from among the connecting lines obtained by the connecting line generating unit and linearly approximating the longest connecting line. A distance between an edge candidate generation unit that obtains side candidates of the rectangular object and a straight line obtained by the edge candidate generation unit among the connecting lines obtained by the connection line generation unit is a predetermined value. an edge element extraction unit that extracts a smaller connecting line; an edge slope candidate detection unit that linearly approximates all the connection lines obtained by the edge element extraction unit and detects the slope of the straight line; and the side slope candidate detection unit. a slope detection section that calculates the slope of the rectangular object from the slope candidates for each side of the rectangular object obtained by the section; and a side detection unit that detects the sides of the rectangular object from all the connecting lines obtained by A contour extraction device characterized by extracting contours.