JPS603073A - Pattern extraction system - Google Patents

Abstract

PURPOSE:To improve pattern extraction by retracing a black picture element string given a specific label, and correcting the label. CONSTITUTION:A labeling device 13 makes a raster scan on the image in a memory 11 from the left upper corner to search for the 1st black picture element. Then, an aimed picture element is moved along outside boundary picture elements so that a clockwise round of a concatenation pattern is made; and thus tracing is performed and label is give. A labeling device 13 traces the labeled picture element string in a memory 11. It is decided in this tracing whether a terminal point is found or not, and when there is no terminal point found, a jump to the next step is made. When the terminal point is found, the label contacting the picture element is corrected and an advance to the next step is made. When the label correction is completed, a segmenting device 14 segments respective patterns of a picture element string with an isolated label and the picture element string with the label from the memory 11 and outputs them.

Description

DETAILED DESCRIPTION OF THE INVENTION [Technical Field] The present invention relates to a pattern extraction method for separating and extracting connected patterns.

[Prior art]

Conventional pattern extraction methods identify the range of individual patterns by determining connectivity on a pixel-by-pixel basis.Therefore, the overall connectivity state could not be grasped, which could lead to cutting errors.For example, In the case of the connected pattern shown in FIG. 1, the line figure 1.2 and the connected line 3 cannot be separated and are extracted as one pattern.

〔the purpose〕

An object of the present invention is to provide a pattern extraction method that can separate and extract the above-mentioned connected patterns.

〔Example〕

An embodiment of the present invention will be described using FIGS. 2 to 7.

FIG. 2 is a schematic block diagram showing an example of a pattern extraction device according to the present invention, in which 11 is a memory, 13 is a labeling device, and 14 is a cutting device. These parts are interconnected by a bus i.

Now, assuming that the image of the connected pattern shown in FIG. 3 has been entered into the menu 11, the pattern extraction process will be explained in order.

FIG. 7 is a schematic flowchart of the process.

First, the labeling device 13 scans the image in the memo I 711 from the upper left corner and searches for the first black pixel (step [phase]). In this example, pixel A is found and becomes the first pixel of interest. A label "2" is attached to the black pixel adjacent to the noticed lli+i element and sog (step 2).

The target pixel is moved by one position along the outer boundary pixel as shown by the arrow in FIG. 4 (step 0). Attach the label "2" to the pixel of interest and the surrounding pixels shown in FIG. 6 (Step 0
). In FIG. 6, the x indicates the pixel of interest, and the arrow shown at the bottom of each figure indicates the direction of movement of the pixel of interest. For example, from pixel to B
++! When the pixel of interest is moved to the Ii element (Figure 3),
The three surroundings lI shown in FIG. 6(b)! Add a label "°2'° to the Ll element and the pixel of interest If the answer is NO (step 0), return to step 0.

In this way, the connected pattern is traced clockwise (by moving the pixel of interest along the outer boundary pixels) and labeled. The result of labeling is as shown in FIG. In this figure, ■ is labeled ``2'' (sara 7
1. This pixel is then relabeled ``3''. In other words, since the connection #40 part is a round trip of tracing, the label "2" is attached on the outward route, but the label "2" is attached on the return route.
3" does not change. In contrast to this 7"L, the line segment shape 5.6 with a spread will not be traced repeatedly.
iJ) and add the label ゛2'' to become Nara n, Tamayo.

As is clear from FIG. 4, the connection line 4 and line figure 5.6 can be extracted separately at this stage as well. However, the label °“3''
If 1 is extracted at this stage, part of the line figure 5.6 will be missing. Therefore, the label of this connected part is corrected.

That is, for labeling, the device 13 traces the 71 element sequence labeled "2'' on the memory 11 (step 0). It is determined whether an end point is found in this trace (step 0), and if there is no end point. If the end point is found, the label "3" touching the white pixel is changed to the label "2". Step [Phase]) and Step [
Phase] Proceed.

To explain with reference to FIG. 4, in the case of line figure 5, Cdii
Trace is made from the line to the C8 pixel, but since these pixels are far apart, there is an end point.

Therefore, the label "3", that is, the C1 pixel and the C2 pixel, which are in contact with the white pixel, are modified to the label "2".

When the label correction of the line segment shape 5 is completed, the process returns from step @ to step 0, and the label correction of the remaining line figure 6 is performed (7). In this case, 0 pixels. Trace to pixel J and l). The pixel is modified to label 11211.

When the label correction is completed as described above, the memory 11
The contents are as shown in FIG. Cutting device 14,
The isolated pixel column with the label "2", that is, the patterns of line figures 5 and 6, and the i[!II purple column of the label "'3", that is, each pattern of the connecting line 4 are cut out from the memory 11 and output to the outside. (Step @).

Note that if the line width of the pattern to be processed is large, the label may not be added normally as shown above. It is desirable to apply

For example, if you perform the operation (reduction/thinning) of calculating the logical sum of 2×2 pixels and converting it into 1 pixel on the image in Figure 3, and then perform the above-mentioned steps [phase] to 0 labeling, , the result is as shown in Figure 8.

It is clear that the same pattern extraction as described above can be achieved by executing the processes from step 0 onwards. The reduction ratio of such an image depends on the accuracy of the original image,
In the case of a handwritten figure with a reading density of 8×8 pixels, the reduction ratio is generally 1/8.

〔effect〕

The present invention will be described in detail below.
Alternatively, it is possible to obtain the effect that similar connected patterns can be separated and extracted.

[Brief explanation of the drawing]

FIG. 1 is a diagram showing an example of a connected pattern, FIG. 2 is a schematic block diagram showing an example of a pattern extraction processing device according to the present invention, and FIGS. 3 to 5 are patterns for explaining IJJ pattern extraction processing. 6 is an explanatory diagram of labeling, FIG. 7 is a schematic flowchart of pattern extraction processing, and FIG. 8 is a pattern diagram showing the result of labeling after reduction and thinning. ]l...Memory, ]2...Bus, 13...Labeled device, 14...Cutting device. Representative Patent Attorney Makoto Suzuki 314 Shi□□ −

Claims (1)

[Claims] (1) While tracing a connected pattern consisting of one line figure and one or more connected lines connected to it along the outer boundary pixels, trace the lines between the black pixels of the line figure and the connected lines. After labeling each black pixel with a different label, the lapel alignment of the connecting portion of the line figure with the connecting line can be determined by tracing again the black pixel row that has been given a specific label. A pattern extraction method characterized by separating and extracting a dashed line figure and the connecting line.