JPS634379A - Pattern matching device - Google Patents

Abstract

PURPOSE:To attain a quick processing even if an image memory has a large capacity by matching patterns macroscopically and microscopically. CONSTITUTION:A block image forming means 2 obtains a block made of picture elements reduced to 1/n (n: the corresponding number of an original image to the block image), compared with its original image. Then a first overlap control means 3 and a macroscopic matching detection means 4 match patterns at a high speed. A feature part specification means 5 specifies a feature part, and a 2nd overlap control means 6 and a fine matching detection means 7 execute the fine matching of the feature part, thereby improving the accuracy of the matching.

Description

[Detailed description of the invention] [Purpose of the invention] (Industrial application field) The present invention relates to a pattern matching device.

(Prior Art) A conventional pattern matching device has an image memory that stores a reference pattern and a comparison pattern, and a cross-correlation coefficient (see formulas );
Pattern matching is detected when the phase n correlation coefficient shows a value that is too large.

(Problems to be Solved by the Invention) However, the conventional pattern matching device uses both image memories to create a finely formed i! Since the structure was such that the two elements were mutually moved and the mutual 11-1 coefficient was calculated each time, there was a problem in that it took an extremely long time to process the bullets.

For example, when attempting to capture an integrated circuit with a scanning electron microscope and store the digital signals obtained by the microscope in an image memory for pattern matching, the number of pixels in the image memory becomes large, making pattern matching difficult. It requires a lot of time.

SUMMARY OF THE INVENTION It is therefore an object of the present invention to provide a pattern matching device that can improve the above-mentioned problems and perform high-speed processing even if the capacity of the image memory is large.

[Structure of the Invention] (Means for Solving the Problems) In order to achieve the above object, in this invention, as shown in FIG. A reference and comparison image forming means 1 is stored in an image memory, and a block image forming means 2 is formed, corresponding to a plurality of pixels in the image memory, to form a block image in which the average density of these pixels is expressed on a unit pixel.
a first superimposition t=iii control means for sequentially superimposing the block images corresponding to the one pattern and the comparison pattern while giving a relative pixel-by-pixel displacement to the block images corresponding to the one pattern and the comparison pattern; A macroscopic matching detection means 4 detects pattern matching between both images by calculating the number of patterns, a feature specifying means 5 specifies a feature having a relatively small area on both original images, A second ff that sequentially overlaps both feature parts while giving a relative displacement in pixel units.
The apparatus is configured to include an i-alignment control means 6 and a precision matching detection means 7 for detecting pattern matching of both feature parts by calculating a cross-correlation coefficient according to the overlay.

(Function) In this invention, the block image forming means 2 obtains block pixels whose number of pixels is reduced to 1/n (n is the number of correspondence between the original image and the block image) with respect to the original image, and
The overlay control means 3 and the macroscopic matching detection means 4 perform pattern matching at a speed n0α times faster than the conventional method. In the magnification n+α, n is a gain that can reduce the number of superimpositions, and α is a gain that can be made simpler.

Further, in this invention, the characteristic part is then specified by the characteristic part specifying means 5, and precision matching is performed on the characteristic part by the second overlay control means 6 and the precision matching detection means 7, so that the matching accuracy can be improved. can be achieved.

Note that the area of the feature specified by the feature specifying means 5 is arbitrarily determined according to the comparison pattern, but since this can be done for a relatively small area, W'' is used for pattern matching of the feature. The amount of time you spend in J is minimal.

(Example) Hereinafter, the present invention will be described in detail using the accompanying drawings.

First, to briefly explain the drawings, FIG. 2 is a block diagram of a pattern matching device according to an embodiment of the present invention, FIG. 3 is an explanatory diagram of an original image, and FIG. FIG. 4(b) is an explanatory diagram showing block images of the comparison pattern and the reference pattern.

Furthermore, FIG. 5 is an explanatory diagram showing the action of the first superimposition control means, and FIGS. 6(a) and 6(b) show the overlap of two block images that are the targets of the calculation of the cross-correlation coefficient. It is an explanatory diagram showing the state of the parts.

Furthermore, FIG. 7 is an explanatory diagram of the concentration gradient that is taken into account when specifying a characteristic part, FIG. 8 is an explanatory diagram of the block position of the characteristic part, and FIG. 9 is an explanatory diagram showing the action of the second overlay control means. be.

As shown in FIG. 2, the pattern matching device is comprised of an image processor 8 having three image screens M 1 , fvl 2, .about.13.

The image memory M1 stores the reference pattern ps as the original reference pattern ii! I will remember it as an image.

The image memory M2 captures the image signal of the comparison pattern Pc via the scanning electron microscope 9 and stores it as an original image of the comparison pattern.

The image memory M3 is a memory used as a working area for the image processor 8.

Image processor 8 uses these memories M1. M2 and M3 constitute each of the means 1 to 7 shown in FIG.

Below, the contents of the processing performed by the image processor 8 will be explained in order.

First, the reference and comparison image forming means 1 shown in FIG.
As shown in FIG.
The original image ES (shown by ) is stored.

Similarly, a comparison pattern Pc is stored in the image memory M2.
The original image EC of is stored.

Storage is performed by storing density data of 256 (0 to 255) levels in each pixel. Here, the data rOJ &Y means black, and the data r255J means white.In this example, in which white punched characters rFJ are shown on a black background, the image memory Ml (M2) has a and a light as shown in the figure. The data will be stored.

There are a plurality of block image forming means 2 (four in this example).
Figure 4 (a), (
As shown in b), the original image @Es (
Block image 0s(cc) with a size of 1/4 of Ec)
Remember.

After performing the above preparatory processing, pattern matching is performed by the cooperation of the first overlay control means 4 and the macroscopic matching detection means 4.

As shown in FIG. 5, the first superimposition control means 3 fixes the block image aS of the reference pattern, for example, and moves the block image iec of the comparison pattern left and right, up and down, pixel by pixel.

- On the other hand, the macroscopic matching detection means 4 (Y, the first middle matching control means 3 has a 1,400-square-meter portion 10 in the middle moving circle of each pixel.
, calculate the cross-correlation coefficient COR.

The cross-correlation coefficient COR is first calculated as shown in Figure 6 (a) and (b).
) As shown in D, the hand shape portion 10c of each block image
.

8 rows and columns matrix (uij) of grayscale data for 10s
, (vij) and form matrices (uij ), (v
The average value of ij > is determined by the following equation, where mu and mv are the standard deviations, and σU and σV are the standard deviations, respectively.

C0R= (1/a -b) ...■ Here, IJ ij= (uij -mu) /au
−■Vij= (vi
j − mv) /σ■ ...■.

The macroscopic matching detection means 4 calculates the above-mentioned cross-correlation coefficient each time in accordance with the pixel-by-pixel movement control of the first overlaying means 3, and considers the maximum value as the suitable state for matching.

From the above, it can be found that the comparison pattern Pc macroscopically matches the reference pattern Ps, and as shown in FIG. 5, the coordinate displacement amount ΔX at this time.

Δy is found.

Next, the feature specifying means 5 performs differentiation on the original image ES of the reference pattern Ps or the original image EC of the comparison pattern Pc, as shown in FIG. 7, to form a density gradient matrix 11 on the memory M3. .

Let dij be the density gradient for the density of pixels in arrays i and j, then dij = d P/dx+d P/dl/=(P!++
, j-Pi, j > + (Pi, J, I PI, J)...
・It is ■.

Also, the characteristic portion specifying means 5 extracts pixels whose density gradient is 100 or more in FIG. 7 as shown in FIG.
The block BLKS in which the area occupied by the portion where ij is 100 or more is the maximum is defined as the characteristic portion.

Further, the characteristic portion specifying means 5 specifies the coordinate displacement old ΔX, Δy
Taking this into account, designate the portion corresponding to blocks B and K shown in FIG. 8 on the original image EC (or Es) of the other pattern as the characteristic portion (referred to as block BLKC) for the other pattern. do.

Therefore, the second superimposition control means 6 controls one of the thus obtained blocks 8LKs and BLKc to the other block, as shown in FIG. 9, in the same manner as shown in FIG. The precise mapping detection means 7 is
In conjunction with this movement, calculations similar to those shown in equations (1) to (2) are performed to perform precise pattern matching.

In order to judge whether the pattern matching has been performed well and it is an IC, it is sufficient to combine the detection results of the macroscopic matching detection means 4 and the detection results of the precision matching detection means 7.

As described above, in this embodiment, pattern matching can be performed at high speed, and, for example, pattern matching of integrated circuits can be performed from 2 to 30 minutes.
This can be done in 3 minutes.

Furthermore, since pattern matching is performed macroscopically and more precisely, it is possible to perform more complete matching processing.

In this example, the characteristic portion specifying means 5 specifies one block with a large concentration gradient, but the characteristic portion may be specified in any manner depending on the pattern type, such as specifying multiple portions, for example.

Further, in this example, one block is made up of four pixels of the original image, but it goes without saying that this number may be other numbers such as 6, 9, 8, 12, 16, etc.

In addition, in this example, the number of pixels in the precise matching shown in Fig. 9 is shown in the same size as the blocks shown in Fig. 4 (, 1), (b) and Fig. 8, but the Dense matching may be performed on specified VT features, for example, in 6 (8 (expanded circles), 9 (expanded angle) shades), or in wide areas.

This invention is not limited to the embodiments shown above, but can be implemented in various ways by making other changes in the currency design.

[Effects of the Invention] As described above, according to the pattern matching device according to the present invention, pattern matching is performed macroscopically and microscopically, so even if the capacity of the image memory is large, high-speed processing can be performed. .

[Brief explanation of the drawing]

FIG. 1 is a diagram showing an overview of the present invention, FIG. 2 is a block diagram of a pattern matching device according to an embodiment of the present invention, FIG. 3 is an explanatory diagram of an original image, and FIGS. FIG. 4(b) is an explanatory diagram without showing block images of the comparison pattern and the reference pattern, FIG. 5 is an explanatory diagram showing the action of the first width alignment control means, and FIG. 6<a) and FIG. b) is an explanatory diagram showing the memory state of the overlapping part of two block images that are the targets of calculation of the cross-correlation coefficient, FIG. An explanatory diagram showing the block position of the characteristic part, FIG. 9 is the second superimposition t
! FIG. 11 is an explanatory diagram showing the action of the means. 1...Reference and comparison image forming means 2...Block image forming means 3...First overlay control means 4...Macroscopic matching detection means 5...Characteristic portion specifying means 6...First 2 Overlay control means 7...Precision matching detection means ps...Reference pattern pc...Comparison pattern

Claims (1)

[Claims] a reference and comparison image forming means for storing a reference pattern and a comparison pattern as original images in an image memory; and a block for forming a block image in which the average density of these pixels is expressed on a unit pixel corresponding to a plurality of pixels in the image memory. an image forming means; a first superimposition control means for sequentially superimposing the block images corresponding to the reference pattern and the comparison pattern while giving a relative displacement in pixel units; A macroscopic matching detection means detects pattern matching between both images by calculating a number, a feature specifying means specifies a feature with a relatively small area on both original images, and a feature specifying means specifies a feature with a relatively small area on both original images. a second superimposition control means for sequentially superimposing both feature parts while giving a unit displacement; a precise matching detection means for detecting pattern matching of both feature parts by calculating a cross-correlation coefficient according to the superposition; A pattern matching device comprising: