JPS63133765A - Pattern detection processing system - Google Patents

Abstract

PURPOSE:To perform the calculation of degree of difference with a simple processing, by detecting the black-and-white change point of a picture signal on one line and the black-and-white change information between lines, and performing the segmentation of an area according to the information of a relative position. CONSTITUTION:A change point information detecting part 2 detects change point information such as the deviation of a change point with a runlength or a preceding line based on the picture signal read from a picture signal memory 1, and set it to a segmentation deciding part 3 and a change point information memory 4. The segmentation deciding part 3 traces inputted change point information along which boundary it is generated, and based on the result of boundary tracing, the change point information memory 4 stores the information independently at every consecutive boundary in a phasewise. Also, when segmentation is decided at the segmentation deciding part 3, the content of the change point information stored in the memory is sent to a difference degree calculating part 5. The difference degree calculating part 5 finds the degree of difference between the content of the information and a fundamental pattern, and generates the same pattern at a deciding part 7.

Description

DETAILED DESCRIPTION OF THE INVENTION (1) Field of the Invention The present invention relates to a pattern detection processing method for detecting a pattern identical to a basic pattern from a single image signal.

(2) Conventional technology Facsimile has become rapidly popular due to its ability to easily send any document regardless of its format. Since the read image signal is encoded as a photographic element, the encoding efficiency is not necessarily good for information that can be encoded as a character element.

As an encoding method for efficiently processing such character information, there is a pattern matching encoding method. This is a pattern that appears within the same document.

In this method, a pattern is registered in a library pattern, and when a similar pattern appears, encoding is performed by sending the rD of that library pattern.

The method of detecting the same pattern used in these methods cuts out the detection area as a rectangle surrounding an isolated black pixel area, and compares the dot pattern in that area with the dot pattern in the library for each pixel. For example, the degree of difference is totaled, and based on this value, it is detected whether or not they are the same.

However, in these pattern detection methods, it is necessary to cut out the isolated black pixel area in a rectangular shape.

In addition, since it is necessary to perform processing such as determining the degree of difference for each pixel, there is a drawback that the amount of processing is enormous. (3) Purpose of the Invention The present invention has been made to solve these drawbacks. An object of the present invention is to provide a pattern detection processing method that can cut out a region for pattern detection and calculate the degree of dissimilarity using simple processing.

(4) Structure of the invention (4-1) Differences between the characteristics of the invention and the conventional technology The characteristics of the present invention are to detect the black and white change point of the image signal on the l line and the black and white change information between the lines. , the region is cut out based on the relative position information, and the degree of difference from the basic pattern is determined.

Therefore, by extracting information along one phase-continuous boundary, it is possible to cut out a region for pattern detection without being a rectangle. Also, relative position information is used to calculate the degree of dissimilarity.

The degree of difference can be calculated by performing processing for each change point, and there is no need to calculate the degree of difference for all pixels within the region. 5 more
Even when image signals are input line by line, the degree of difference can be calculated for each line. Further, since the relative position information of the change point is used in MR encoding used in facsimile, if the MR encoding information is used, it is compatible with facsimile and the like.

(4-2) Example 〔Example! ] FIG. 1 shows a first embodiment of the present invention.

In the figure, 1 is an image signal memory in which image signals for which the same pattern should be detected are stored. Reference numeral 2 denotes a changing point information detecting section, which detects changing point information such as run length and deviation of a changing point from the previous line based on the image signal read out from the 5-image signal memory l.

Reference numeral 3 denotes a cutout determination unit, and the change point information obtained by the change point information detection unit 2 is 1. For example, in FIG. 2, thick lines (a) to (d)
With regard to the boundaries shown in , trace the boundary along which the occurrence occurs, and determine the tracking results and the extraction of the area, which is the unit for pattern detection, according to rules such as when an isolated area is detected, It outputs two results including the judgment result.

Reference numeral 4 denotes a changing point information memory, which stores the information obtained by the changing point information detecting section 2 separately for each of nine topologically continuous boundaries based on the boundary tracking result of the cutout determining section 3. In the figure.

Regarding the boundaries of (a) to (d), change point information is individually memorized. Further, when the extraction determining unit 3 determines that extraction is necessary, the memory contents of the change point information generated along the extracted boundary are sent to the dissimilarity calculating unit 5.

Reference numeral 5 denotes a dissimilarity calculation unit, and when the extraction judgment unit 3 determines whether to cut out and the change point information is sent from the change point information memory 4, the information content and the basic pattern library 6 are stored for each basic pattern. The degree of difference from stored information is determined separately for each basic pattern. 7 is a judgment section;
From the dissimilarity calculated by the dissimilarity calculation unit 5, a basic pattern that is the same as the area cut out according to the 1 judgment rule is detected.

The determination result is output to the output terminal 8.

Next, the operation will be explained using a simple example.

The changing point information includes various things such as the size and position of the deviation of the changing point from the previous line, the number of deviations, the direction of the deviation, etc. Here, as an example of the changing point information, the beginning of the black pixel area The case of using the black run length of , and the deviation from the previous line regarding the changing point of the left and right boundaries in the black pixel area will be described with reference to FIG.

First, the image signals are sequentially examined to detect the beginning of a black pixel area. For the determined beginning, the black run length R is determined. In the example of FIG. 2, it is determined as *R''2''1.

Next, the shift point Δ” at the left and right boundaries of the black pixel area
(1) + a, , bz at) are found one after another for each line. However, +al+ b+ is from white pixel to black pixel + a! + b! is the change position from a black pixel to a white pixel.

Examples of extraction of change point information for the boundaries (a) to (d) shown in FIG. 2 are shown in FIG. In the basic pattern library 6 shown in FIG. 1, change point information for basic patterns is similarly extracted and stored as a library.

Assume now that the image signal shown in FIG. 4 is input.

The change point information regarding the boundary N) is obtained as shown in FIG. 5, and the degree of difference between this information and the contents of the basic pattern library 6 is determined. Assuming that the information for is stored as a basic pattern, the degree of difference is determined between the change point information in FIG. 3 and the change point information in FIG. 5.

There are two different methods for calculating the difference in change point information, but here we calculate the absolute value of the difference in each change point information between patterns with the same boundary height H, and calculate the sum as the difference degree. In this example, we will explain the case where:

(a) and (d) match the height of the boundary (+) in Figure 4,
The degree of difference is between (a) and (i), respectively, and ABS (
Ri Ra)+ABS (Δdirect, -Δ,)+・・・
...=Q+Q+0+l+1+o+o+t+o+o+100-4, (d), (0+1+1+1 between +>
+3+O+O+1+2+O+5+1=15. As a result, (i) is detected as the same pattern as (a) with a small degree of difference.

Regarding the calculation of the degree of dissimilarity, of course it may be defined when the heights of the boundaries are different, or the correspondence of the change points for calculating the degree of dissimilarity may be changed.

Other dissimilarity criteria may also be used.

Regarding the determination of whether or not they are the same, if there is no pattern with a small degree of difference, it may be determined that no identical pattern is found, or the pattern may be newly registered in the library.

[Example ■]

FIG. 6 shows a second embodiment of the present invention.

Mod i is an international standard encoding method for black and white binary images.
fied READ (MR) method (CCITT,
The case where Recommendation T, 6) is used is shown.

In the figure, reference numeral 10 denotes an image signal input terminal, into which one image signal is input line-sequentially. 11 is 2 for extracting change point information.
This is image signal memory for lines.

The image signal is sent to 12 MR encoders and change point information is extracted. The extracted information is sent to the extraction determination section 13,
Boundary tracking is performed.

Further, the difference calculation unit 14 successively compares the contents with the contents of the 5 basic pattern library 15 to calculate the difference. 16
is a dissimilarity memory, which stores the dissimilarity calculated by the dissimilarity calculation unit 14 based on the results of the extraction determination unit 13 separately for each boundary and for each basic pattern, and stores the dissimilarity calculated by the dissimilarity calculation unit 14 based on the results of the cutout determination unit 13, separately for each boundary and for each basic pattern, and stores the dissimilarity calculated by the dissimilarity calculation unit 14 based on the results of the cutout determination unit 13. Each time , the sum is calculated with the memory contents and the contents are rewritten.

The determination unit 17 sequentially performs determination according to the memory contents of the dissimilarity memory 16. Even in the middle of the boundary, patterns whose dissimilarity exceeds a specified value are excluded from the target of the same pattern, and the same patterns are selected from the remaining patterns until the end. seek. What is the judgment result?

In this embodiment, pattern detection processing can be executed sequentially.

〔Example! ], The contents of the basic pattern used in [Example 2] may be universal in advance, or may be determined from a given manuscript. Furthermore, when a pattern whose degree of difference is equal to or greater than a reference value is generated, the library may be updated sequentially, such as by registering that pattern as a new basic pattern. Furthermore, the area in which pattern detection is performed.

It may be forcibly cut out in the middle of the boundary according to predetermined rules.

(5) As described in detail, according to the present invention, it is possible to cut out the area for pattern detection without making it rectangular, and it is necessary to calculate the degree of dissimilarity for all pixels in the area. , the processing is simplified and can be performed at high speed.

[Brief explanation of the drawing]

FIG. 1 shows a first embodiment of the present invention. FIG. 2 is a diagram for explaining detection of change point information. FIG. 3 is a diagram showing change point information for the illustrated boundaries <a> to (d) in FIG. FIG. 4 is an example of an image signal in which the same pattern is detected. FIG. 5 is a diagram showing change point information for the illustrated boundary (1) in FIG. 4. FIG. 6 shows a second embodiment of the invention. l... Image signal memory 2... Change point information detection section 3
... Extraction judgment section 4 ... Change point information memory 5.
...Difference calculation unit 6...Basic pattern library 7...Judgment unit 8...Output terminal IO...
Input terminal 11... Image signal memory 12... MR
Encoder 13... Extraction determination section 14... Dissimilarity calculation section 15... Basic pattern library 16... Dissimilarity memory 17... Judgment section 18... Output terminal

Claims (2)

[Claims] (1) In a pattern detection processing method that detects a pattern corresponding to a basic pattern from various black and white patterns occurring in an image signal, black and white transition points of the image signal are detected within and between lines, and the phase means for determining relative position information of changing points occurring along a continuous boundary; means for storing relative position information of changing points of the basic pattern; and means for storing relative position information of changing points of the basic pattern and the determined means for comparing the relative position information of the change point and calculating the degree of difference between the two; and determining whether the area surrounded by the topologically continuous boundary is the same as the basic pattern based on the calculated degree of difference. 1. A pattern detection processing method, comprising means for detecting. (2) As a means for obtaining relative position information of the change point,
The pattern detection processing method according to claim (1), characterized in that change point information obtained by MR encoding is used.