JPH0296286A - Partial picture area separating device - Google Patents

Abstract

PURPOSE:To quickly separate partial picture areas by newly providing an adjacent edge labeling part, an edge outside margin labeling part, and an adjacent closed area extracting part and permitting picture areas to be separated to be adjacent to each other. CONSTITUTION:An edge extracting part 1 uses density information or the like to extract the outline of a separation object area. An adjacent edge labeling part 2 and an edge outside margin labeling part 3 are used to give a label (e) to the edge of an area A and to give label '0' to picture elements on the outside of the edge of the area A. When another separation object area B exists adjacently to the area A, an adjacent edge label (d) is given to a part adjacent to the edge of the area A. The labeling part 3 gives label 'double circle' to picture elements on the outside of the area having the adjacent edge label. An adjacent closed area extracting part 4 traces the outside of the edge clockwise and extracts one column of picture element groups from the picture element to the boundary picture element in the horizontal direction from the left to the right during advance along the left edge.

Description

DETAILED DESCRIPTION OF THE INVENTION Field of the Invention The present invention relates to an apparatus for automatically separating and extracting a part of a figure or image for recognition or the like.

Conventional technology In recent years, with the advancement of practical applications such as OM and FA, the need to recognize figures and images has increased, but as pre-processing,
From the input digital figure/image information (hereinafter referred to as image information), only the image of a partial area to be recognized, such as a character area, is separated, and the coordinates and pixel values of the constituent pixels of the partial area are separated. and circumscribed rectangle frame information, etc.
The importance of devices that extract information from partial image regions is also increasing.

Hereinafter, two sides of the conventional partial image region separation device will be described with reference to the drawings.

FIG. 5 is a block diagram showing the configuration of a first example of a conventional partial image region separation device. In FIG. 6, 1 is an edge extraction section, 7 is an edge labeling section, 8 is a closed region extraction section, 6
6 is an image storage unit, and 6 is an image area information storage unit.

The operation will be explained below. First, an edge extraction section extracts the outline of the region to be separated, that is, the edge, using information such as pixel density differences, color information, or higher-resolution image information. Next, each pixel constituting the edge is given a label (pixel value) indicating the edge, as shown in FIG. In Fig. 6, “○” is a label representing an edge constituent pixel, and “°′・” is a label representing a pixel other than an edge in the image area to be separated. There is no need to have the same label as in -.Also, pixels outside the image area to be separated only need to have a label other than "101", and may be the same as the label of the pixels in the above area. .

An image labeled in this way is stored in the image storage unit, and when it is desired to obtain a target partial image region and its information, the closed region extraction unit extracts information on edges and pixels inside the edge and extracts the information from the image. Store in area information storage unit.

Further, FIG. 8 is a diagram showing the configuration of a second example of a conventional partial image region separation device. In FIG. 8, reference numeral 9 denotes an area extraction section, 10 an area labeling section, 11 a labeling area extraction section, 5 an image storage section, and 6 an image area information storage section.

The operation will be explained below. First, pixels belonging to the region to be separated are determined using various types of information such as those used in the edge extraction section of the first example. Next, all pixels constituting the area are given labels unique to that area. FIG. 9 shows a state in which two regions to be separated are labeled when they are adjacent to each other. Pixels with the label ``Pa■'' belong to one area, and pixels with the label 1■'' belong to the other area.

An image labeled in this way is stored in the image storage unit, and when it is desired to obtain a target partial image area and its information, the labeling area extraction unit extracts the information of the pixels having the label of that area. Store in the image area information storage unit.

Problems to be Solved by the Invention However, the above-described configuration has a problem in that it cannot process large-scale and complex images, which have come to be handled in recent years, at high speed and economically.

In other words, in the first conventional example, if two or more regions to be separated and extracted are adjacent as shown in FIG. It is difficult to correctly extract edges in each region. Furthermore, in order to avoid this problem, if the labels given to the edges are unique for each partial image region to be separated and extracted, the types of labels given will also increase when the number of regions to be separated increases. Since the amount of information required for labeling also increases, the storage capacity required per pixel must be increased, and the capacity of the entire image storage unit must also be increased. Or the seventh
After labeling, separating, and extracting the region people in the figure, restore the edge labels, and then label, separate, and extract region B.
Although it is possible to extract the images and information of a person from another area, it is difficult to store the images and information of all partial image areas (using a huge storage capacity). (If you do not store the edges in a file, you will have to start over from the first edge extraction.)

In the case of the second conventional example, the adjacent area problem and the reprocessing problem do not occur, but similar to the problem described in the first example above,
As the number of regions to be separated increases, a large-capacity image storage unit is required, and since all pixels within the region are labeled, not just edge pixels, processing in the region labeling unit takes a lot of time. There is a problem with this.

In view of the above-mentioned problems, the present invention is capable of correctly separating and extracting each region even if multiple image regions to be separated are adjacent to each other, and it is possible to correctly separate and extract each region from an image labeled with a - degree. The present invention provides a partial image region separation device that is capable of separating and extracting information, does not require a large capacity of an image storage unit, and is capable of processing at high speed.

Means for Solving the Problems In order to solve the above problems, the partial image region separation device of the present invention includes an edge extraction unit that detects edges (contours) of image regions to be separated, and a plurality of image regions to be separated that are adjacent to each other. an adjacent edge labeling unit that assigns a pixel value to the edge constituent pixel in consideration of the case where the edge constituent pixel is an adjacent closed region extraction unit that separates and extracts an image region to be separated from pixel values of pixels adjacent to the outside; an image storage unit that stores pixel value information of the image;
and an image area information storage unit that stores information regarding the separated and extracted partial image areas.

Effect of the Invention With the above-described configuration, the present invention performs labeling in consideration of the adjacency state between a plurality of separation target regions in the adjacent edge labeling section and the edge outer edge labeling section, while maintaining processing speed comparable to that of the conventional method. By cutting out the inner region of the image labeled as above in the adjacent closed region extraction section, each region to be separated does not have its own label, so it can be used multiple times without requiring a large capacity in the image storage section. Separate and extract partial image regions that can be processed.

Embodiment Hereinafter, a partial image region separation device according to an embodiment of the present invention will be described with reference to the drawings. FIG. 1 shows the configuration of an embodiment of the present invention. In FIG. 1, 1 is an edge extraction section, 2 is an adjacent edge labeling section, 3 is an edge outer edge labeling section, 4 is an adjacent closed region extraction section, 6 is an image storage section, and 6 is an image region information storage section.

The operation will be explained below. Although the digital image is assumed to be in the form of a square grid, it is possible to operate in the same way with a hexagonal grid or the like with slight changes.

The edge extraction section is the same as that in the first conventional example, and extracts the outline of the image area to be separated using density information, high-resolution image information, and the like. Next, the manner of labeling by the adjacent edge labeling section and the edge outer edge labeling section will be explained using FIG. 2. First, a label "°°■" representing the edge is given to the edge of the region M. Then, to pixels outside the edge of the region M (pixels adjacent to the outside of the edge in 8-connection), a label representing the outer edge is given to the edge of the region M. Give the label °゛○”. if,
If the large region exists alone (with no other adjacent regions to be separated), the shape remains as is, and separation and extraction can be performed in the same manner as in the first conventional example. As shown in FIG. 2, when there is another separation target area B adjacent to the large area, the following will occur. The adjacent edge labeling part is
In fact, it does not just mechanically give a label of "■" to an edge pixel. The type of label to be given is determined by looking at the pixel value (label) of the edge pixel before labeling. In Figure 2, A part of the edge pixels of area B (adjacent to the edge of area B) is initially labeled with a label "O'" indicating the outer edge (when labeling the area size). In this case, it is adjacent to the edge of another separation target area (area A in this case), so the label to be given is the adjacent edge label "@".At this time, in the edge outer edge labeling part, the edge of the adjacent edge label is A label "■" representing the outer edge of the adjacent edge is given to the 8 connected pixels outside the area.However, even if the edge outer edge labeling part is a pixel at the outer edge of the edge, if it is an edge pixel in another area, the label is added. Therefore, the edge of the region M is not cut in Figure 2.If the edge is diagonal and there is an edge of another region to be separated adjacent to it, The inner edge of a region labeled as an adjacent region may sometimes be labeled as the outer edge.The adjacent edge labeling unit detects such pixels and restores the label to its original value. This corresponds to the pixels marked with ``x'' in area B. And vice versa,
Pixels inside an already labeled region may become the outer edge of another adjacent region. When the edge labeling section detects such a pixel, it does not attach a label representing the outer edge. This corresponds to the pixel marked 9Δ'' in the area M in Figure 2. In this way, the adjacent edge labeling section and the edge outer edge labeling section are labeled while also referring to the labels of surrounding pixels. In the case of a digital image in the form of a square grid, 1
Since up to four areas may be adjacent to a point, triple and quadruple labels for edges and outer edges are prepared. When the original label of an edge pixel is the outer edge "■" of an adjacent edge, a triple adjacent edge label is given, and the outer edge pixel is given a label representing the outer edge of the triple adjacent edge. The same thing applies when there are 4 layers. If the digital image is in the form of a hexagonal grid, up to three labels are sufficient. In actual labeling, labels can be given to edge pixels and pixels on the outer edge at the same time (in parallel), so there is no loss in speed.

By storing images labeled in this manner in the image storage section, it becomes possible to separate and extract images and information any number of times using the adjacent closed region extraction section. The adjacent closed region extraction unit tracks the outside of the edge in a clockwise direction, and while moving along the edge on the left side (that is, while the vertical component of the advancing direction is pointing upward), it extracts the boundary pixel horizontally to the right from that pixel. Extract a row of pixels up to. The boundary pixel is the pixel one pixel to the left of the edge pixel of a different type from the edge pixel at the starting point, and also the pixel one pixel to the left of the edge pixel when it hits a pixel representing an outer edge of any type. becomes. If this is performed while tracing the image along the edge of the area until it goes around (until it returns to the starting point), it becomes possible to separate and extract all the pixels within the area. The pursuit of Etsuji is
If there is an edge pixel of the same type in a direction other than the edge pixel at the point one pixel before (the pixel in the original direction in which it progressed) within the 8 pixels in the 8 connected directions of the edge pixel of interest, move in that direction; This is possible by moving in the direction of a different type of edge pixel if there is one, and if there is none, returning in the direction of the previous pixel. However, as shown in Figure 4, the protruding pixels (this is area C
If there is a pixel in the area), it would be a problem if it goes in the direction of the edge pixel of the area, so such protruding end point pixels should be labeled in advance (because they can be easily detected). '', and when proceeding to a pixel with this label, it will unconditionally return to the original direction.Adjacent edge pixels and triple/quadruple adjacent edge pixels are also Prepare similar end point labels.These end point labels, edge pixels, adjacent edge pixels, etc.
Regarding edge tracking, it is assumed that it is the same type (otherwise, endpoint pixels will be left unmarked).

In this way, by simply adding the above 12 types of labels (9 types for hexagonal grid images) to some pixels, each area can be correctly separated and extracted even if there are adjacent image areas to be separated. be able to.

Note that when giving a label to each pixel, we will rewrite the pixel value, but if you want to retain the original pixel value information, you can add each of the above labels to the pixel value as an offset. You can take it.

Effects of the Invention As described above, the present invention provides a new arrangement of an adjacent edge labeling section, an edge outer edge labeling section, and an adjacent closed region extracting section, thereby making it possible to detect that image regions to be separated are adjacent to each other. It allows for multiple processing, does not require a large capacity in the image storage unit, and enables high-speed partial image region separation. It provides processing.

[Brief explanation of the drawing]

Fig. 1 is a block diagram of a partial image region separation device according to an embodiment of the present invention, Fig. 2 is a data diagram showing an example of the result of image labeling by the above device, and Fig. 3 is a diagram of adjacent images on a digital image. A data diagram explaining the number of regions, FIG. 4 is a data diagram showing an example of the result of labeling an image with a protruding part by the above device, and FIG. 5 is a configuration of the partial image region separation device of the first conventional example. 6 is a data diagram showing an example of the result of labeling an image using the same device, FIG. 7 is a data diagram showing the problem of labeling using the same device, and FIG. 8 is a partial image of the second conventional example. FIG. 9 is a block diagram of the region separation device, and is a data diagram showing the result of labeling an image by the device. 1... Edge extraction section, 2... Adjacent edge labeling section, 3... Edge outer edge labeling section, 4... Adjacent closed region extraction section, 6.・・・・・・
Image storage unit, 6... Image area information storage unit, 7.
... Edge labeling section, 8... Closed region extraction section, ... Region labeling section, Placement section.

Claims (1)

[Claims] an edge extraction unit that extracts edges of a separation target image; an adjacent edge labeling unit that assigns pixel values to edge constituent pixels in consideration of the case where multiple separation target image regions are adjacent; and an adjacent edge labeling unit that assigns pixel values to edge constituent pixels; an edge outer edge labeling unit that assigns pixel values to pixels adjacent to the edge, an adjacent closed region extraction unit that separates and extracts an image region to be separated from the pixel values of the edge constituent pixels and pixels adjacent to the outside, and a pixel value of the image. A partial image region separation device comprising: an image storage section that stores information; and an image region information storage section that stores information about separated and extracted partial image regions.