JPS6069783A - Picture collation system - Google Patents

Abstract

PURPOSE:To speed up pattern matching processing such as seal impression collation by using the density sum of picture elements present on a radial half-line passing the center of rotation of an picture as a feature parameter. CONSTITUTION:A picture having a contour of symmetry of rotation is an object of collation and the coordinates (X0, Y0) of its center of rotation are calculated from equations 1 and 2. Then, the number of black picture elements on the half- line which originates from the center (X0, Y0) of rotation and has an angle theta(0<=theta<=2pi) to an X axis is represented as f(theta) and called a density pattern. When the density patterns of a registered picture and the picture to be collated are f1(theta) and f2(theta), the angle phi of rotation required to superpose both pictures one over another is calculated from equations 3-6. Therefore, calculations are simplified as compared with a system which uses a density distribution on concentric circles or in a sector area, and pattern matching processing is speeded up.

Description

DETAILED DESCRIPTION OF THE INVENTION [Field of Application of the Invention] The present invention relates to an image matching method, and mainly relates to an image matching method suitable for pattern matching processing of an image having a rotationally symmetrical outline such as a seal image.

[Background of the invention]

There is a growing demand for automation of image verification such as seal verification mark identification. To talk about seal verification as a typical example, seal verification refers to a registered image of a seal filed in advance and a second seal image (verification image) read from an input device.
The main focus is so-called pattern matching processing, in which the two images are superimposed and the degree of agreement between the two is evaluated.

FIG. 1 is a conceptual diagram of an image matching system to which the present invention is applied. From the registered images stored in file 5,
Extracts the corresponding image data according to the specified image name, and
A matching image 1 is read from the image input device 2, and pattern matching processing is performed on both image data. At this time, the processing device 3 translates, rotates, and superimposes the images to create a
Evaluate the degree of matching between both image data. In the image matching system targeted by the present invention, it is assumed that all image data handled has a rotationally symmetrical shape (for example, a circle, an ellipse, a rectangle, etc.). In order to realize the image matching system shown in FIG. 1, a technique is required to align both images with high precision and at high speed.

Several methods have been proposed in the past for this positioning technique. FIG. 2 shows an example of this, in which the density distribution 7 on a concentric circle about the rotation center 6 of the image object is used as a feature parameter, and the density distribution 7 on the concentric circle of the registered image and the verification image is used. This method is based on the song and calculates the rotation angle for superimposing both images. Figure 3 shows another example of the conventional method, in which an image is divided into a large number of fan-shaped areas 9 by a radial straight line passing through its center of rotation g (t), and the headrest on each fan-shaped area is It is a method that uses values as characteristic parameters (Kaneko, Yakushu; Alignment method of seal imprint patterns based on the correlation of peripheral density with respect to the centroid, Materials of the Institute of Electronics and Communication Engineers Research Group,
IE81-13).

These conventional methods require complicated coordinate calculations to read the image data from the image memory because they calculate the density distribution on concentric circles or fan-shaped areas.
It has been difficult to achieve high-speed matching processing.

[Purpose of the invention]

An object of the present invention is to solve the problems of the prior art and to realize pattern matching processing such as seal stamp verification at high speed.

[Summary of the invention]

The basic principle of the present invention will be explained below. However, for convenience of explanation, pattern matching processing for binary images will be described as an example.

In the present invention, an image having a rotationally symmetrical contour is targeted, and the coordinates (Xo, Yo) of the center of rotation are calculated using the following equation.

Xo-(X, nIn + X mali/2 ”(1)Y
o - (Yli + n + Y m-x) / 2 - (2)
Here, X min + X max * Ymln
! As shown in FIG. 4, Y max represents the coordinates of four points in a rectangle that has four sides parallel to the coordinate axes and circumscribes the image. The division formula for the coordinates of the center of rotation is the above equations (1) to (2).
), the applicant, etc. had already proposed Patent Application Sho (5i'-J
3-θ7/) may be used.

Next, calculate the number of black pixels (pixels with a density of 1") that exist on a half-line that forms an angle θ (0≦θ≦2π) with the X axis, starting from this rotation center (Xo, Yo'). Symbol f(
θ) (see Figure 5). It is defined for each θ in the range 0≦θ≦2π, and this is called a “density pattern.” When the density patterns of the registered image and the verification image are defined as fl(of, fl(of), the density patterns of both images are The rotation angle ψ required for overlapping is calculated using the following formula.

cp −Max −” C(9+) −(3)ψ Here, C(ψ)=12″(fl(me−μm)X(fl(θ+ψ
) - μ2) dθ... (4) μl -f fl(dθ/2π... (5) μ2
−fl“fl(dθ/2π (6).

That is, Equations (3) to (6) mean that the rotation angle ψ is determined so that the cross-correlation between the density patterns of the registered image and the verification image is maximized.

Note that the rotation angle ψ can also be determined by using the following equation, which is a much simpler calculation equation, in place of equations (3) to (6).

p = Min-” D (?) −(7)ψ Here, D(ψ)-fl”1f1(-f2(θ0ψ)ldθ
...(8).

The above density pattern f (is, as shown in FIG. 5,
This represents the number of black pixels existing on a radial half line passing through the rotation center of the image. Therefore, compared to the conventional method which uses density distributions on concentric circles or fan-shaped regions, calculation is simpler and pattern matching processing can be performed at high speed.

[Embodiments of the invention]

Hereinafter, the present invention will be explained in detail using examples.

FIG. 6 is a block diagram of an image matching system according to the present invention. In the figure, 10 is a processing device such as a microprocessor, 20 is a main memory, 30 is an image memory, 40 is a keyboard, 50 is an image input device, 60 is a file, and 70 is a display. Image data of registered image (binary image)
is input in advance using the image input device 50,
It is assumed that the image data is stored in the file 60 together with the extracted density pattern. In order to store management information of registered images, a registered image management table 80 shown in FIG. 7 is stored in the main memory 2.
Hold it above 0. This registered image management table 80 includes a 1fiI image name 81 for all registered images, a storage address 82 in the image data file 60, and a density pattern file 6.
Storage address 83 at 0, (x, y) coordinates of rotation center 8
This is a table that stores the 4th prize.

The processing flow of the processing device 10 is shown in FIG. The processing contents of each step in FIG. 8 are as follows.

(a) Geography 110: The command to be executed and the name of the registered image to be processed are taken in from the keyboard 40.

(b) Process 120: 'Decode the input command, and if it is an end command, end the process and stop the operation of the device. If it is not an end command, the process advances to step 130.

(C) Process 130: Find the image data storage address, density pattern, and (X, Y) coordinates of the rotation center for the designated registered image from the registered image management table 80.

(d) Process 140: Read the image data and density pattern of the specified registered image from the file 60. The read image data is stored in the image memory 30, and the density pattern is stored in the main memory 20, respectively.

(e) Process 150: Read the verification image on the input document from the image input device 50,
It is stored on the image memory 30. The registered image is stored in a separate area from the storage area for the registered image.

(f) Process 160: The coordinates (Xo, Yo) of the rotation center of the matching image are calculated using the formula (
Calculated by υ~cut).

(Note) Process 170: Calculate the density pattern f( of the verification image. The method of calculating f() will be explained with reference to FIG. 9. In FIG. It is expressed as a set.A point 210 indicates the rotation center of the matching image.A straight line 220 passing through the rotation center 210 and making an angle θ with the X axis is shown.Starting from the point 210, on the straight line 220 The verification image is sampled at a constant interval a, and the sum of the densities at these sampling points (marked with * in Figure 9) is the density minus the turn f (.So, the sampling points on the straight line 220 are generally , does not match the lattice point (0 mark) of the matching image, so it is necessary to associate the sampling point with the lattice point.
Suppose that the sampling points m=1+2+...,M) are associated with the lattice points in the i-th row and j-column defined by the following equation.

i-C((m-1)XΔX+Xo)/a:)-(9
) j−c ((m−i) XΔY×Yo )/a )−
(to) daso, Δx=acosθ −(11) ΔY=, ll sinθ...(It is effective, the symbol [] means truncation.

Now, the density of the grid point in the i-th row and J column of the matching image is denoted by symbol g1.
When expressed as j, gij takes a value of 0 or 1 since this embodiment deals with binary images. At this time, the density pattern f is calculated using the following equation.

f(-Σ gij m++1 Here, M represents the number of sampling points on the straight line 220.

(11) Process 180: Determine the rotation angle ψ of the matching image necessary to superimpose the registered image and the matching image. The formula for calculating ψ is formula (3) ~
(6) or formulas (7) to (8) are used.

(i) Process 190: The verification image is translated in parallel and rotated by an angle ψ so that the rotation centers of the registered image and verification image coincide, and the two images are superimposed. As a result of the matching, a portion of both images that becomes clear and a mismatched portion are distinguished and displayed on the display 70. Although one embodiment has been described above, the idea of the present invention can be realized by either software or nodeware. In particular, in the case of hardware implementation, in addition to the above embodiment, it is also possible to implement a dedicated processor for the e-degree matching shown in FIG. 9, and in this case, the speed will be even higher.

〔Effect of the invention〕

As explained above, the present invention uses the sum of concentrations of pixels existing on a radial half-line passing through the center of rotation of an image as a feature parameter, and uses the concentration distribution on a concentric circle or fan-shaped area. Compared to the conventional method, since the calculation is directrix, it is possible to speed up the image matching process (pattern matching process).

In the above embodiment, for convenience of explanation, a binary image was described, but the principles of the present invention can also be applied to a multilevel image or a color image.

[Brief explanation of drawings]

Fig. 1 is a conceptual diagram of an image matching system to which the present invention is applied, Fig. M2 and Fig. 3 are explanatory diagrams showing feature parameters for matching one image in the conventional method, and Figs. This is a detailed explanation of the invention, and is an explanatory diagram showing a method for calculating the rotation center of an image and a density pattern on a radial half line passing through the rotation center. FIG. 6 is an illustration of an image matching system showing an embodiment of the invention. A block diagram, FIG. 7 is a configuration diagram of the registered image management table stored in the main memory of FIG. 6,
FIG. 8 is a processing flowchart of the processing device in FIG. 6;
The figure is an explanatory diagram showing a density pattern calculation method in the present invention. 10... Processing device, 20... Main memory, 30...
Image memory, 40...keyboard, 50...image input device, 60...file, 70...display. Figure l ↑ Big Mouth Figure 2 Figure 36 4th Mouth 5 Figure 6 Mouth 8th Mouth 0 0 0 0 0 0000 o o o o o, J6' Claw o, -1 20 000 000T α 000 Sat Oo 0

Claims (1)

[Claims] 1. Calculating the density distribution harameter of pixels along a radial half-line passing through the center of rotation for an image having a rotationally symmetrical contour, and an image required for overlapping a plurality of images. An image matching method comprising the step of determining a rotation angle of the image. 2. A device for comparing images having rotationally symmetrical contours, comprising means for calculating a set harameter on each of a plurality of half-lines passing through the rotation centers of the images to be compared and the images to be compared; An image matching method comprising means for calculating the difference in rotation angle between the two images based on the density parameter, and means for matching the centers and directions of the two images in the closest state.