JPS60116076A - Extracting circuit of picture - Google Patents

Abstract

PURPOSE:To reduce extraction of a background and noise and to extract a recognized object picture precisely by detecting a singular area in an original picture and by selecting elements in the prescribed density range successively connected to the area. CONSTITUTION:In terms of an original picture of a picture memory 1, projection distribution for an (x) co-ordinate and a (y) co-ordinate is obtained by a projection circuit (x) 2 and a projection circuit (y) 3, and stored in two units of memories 4 and 5. A detection circuit 6 detects a singular area in correspondence to areas (DELTAx and DELTAy) in which elements more then a prescribed value in the projection distribution continue over the prescribed range, elements in the singular area are discriminated in a discrimination circuit 8, and the result is stored in a memory 10. An original picture in the picture memory 1 is connected to the memory 10 by a discrimination circuit 9, elements having a density over the prescribed value are discriminated and stored in the memory 10. This process is repeated, and the elements excluding a frame of a background and noises are stored in a picture memory 12.

Description

[Detailed description of the invention] (a) Technical field of the invention.

The present invention relates to image processing, and particularly to an image extraction circuit used to separate only a desired image from a background with noise or the like.

(b) Background of the technology For example, a seal verification system, which is an application field of image processing, observes the stamp to be verified that is stamped inside a specified frame, along with the background that includes the frame, using a method such as raster scanning. For example, each density is represented by a 4-digit binary number for each pixel of 0.1 mm square.
an observation unit that converts into 6-level multi-valued data; an image extraction circuit that extracts only the image to be compared by removing the frame and dirt on the background from the original image obtained by the observation unit; The basic components include a matching dictionary that stores a recognition reference pattern, and a matching section that matches the pattern of the matching target image imaged by the image extraction circuit with the reference pattern stored in the matching dictionary. .

(C) Prior art and problems Conventionally, as an image extraction circuit used for the above-mentioned applications, an original image obtained by an observation unit and represented by multivalued data is binarized with a predetermined value as a dark value, and There is also a method that extracts only "0" pixels.

However, according to such a circuit, background or dirt (
When the density of noise (noise) is close to the density of the image to be compared, such background and dirt are extracted as part of the image to be compared, resulting in a reduction in the recognition rate in the recognition unit.

(d) Purpose of the Invention An object of the present invention is to provide an image extraction circuit that can accurately extract a matching target image even when the density of the background or dirt is close to the density of the matching target image. .

(e) Configuration of the invention, that is, the image extraction circuit according to the present invention is a projection system that obtains a projection distribution of pixels in a predetermined density range in an original image in which the density of each pixel arranged in a matrix is expressed by multi-value data. a detection circuit for detecting a singular region in the original image from a projection distribution obtained by the projection circuit; The above object can be achieved by including an identification circuit that identifies pixels, and a selection circuit that selects from the original image a pixel that corresponds to the pixel identified by the identification circuit.

(f) Examples of the Invention The gist of the present invention will be specifically explained below based on the implementation side.

FIG. 1 shows a professional drawing of the present invention-embodiment, and 1 shows the recognition target seal imprinted inside a designated frame on a sheet of paper along with the background including the frame. The density of each 0.1 mm square pixel is measured in 16 steps (O
-15) An image memory that stores the original image obtained by converting it into multivalued data.

2 and 3, the projection distribution of pixels with density other than 0 (1 to 15) in the original image stored in the image memory 1 to the horizontal (X) axis and the vertical (y) axis is tj? projection circuit, 4
and 5 are memories for storing the projection distributions fied by the projection circuit 2 and the projection circuit 3, respectively; 6 is a detection circuit for detecting a unique region in the original image from the projection distributions stored in the memories 4 and 5; 7 is a memory for storing the singular region detected by the detection circuit 6; 8 is an identification circuit for identifying pixels of the peculiar region obtained by the detection circuit 6 in the original image stored in the image memory 1;

Reference numeral 9 denotes an identification circuit for identifying pixels of the predetermined density or higher that are identified by the identification circuit 8 and sequentially connected to the saw pixels in the original image stored in the image memory 1; Record the position of the pixel in 1. 11 is a selection circuit that selects a pixel corresponding to the pixel identified by the identification circuit 8 and the selection circuit 9 from the original image stored in the image memory 1; 12 is a memory for storing the image selected by the selection circuit 11; It is an image memory that

In FIG. 2, (al) indicates an example of an original image stored in the image memory 1, ■ indicates a frame provided on the paper to designate a sealing area, and ■ indicates an image to be recognized.

The projection distributions obtained by the projection circuit 2 and the projection circuit 3 for an original image such as FIG. 2 (al) stored in the image memory 1 are shown in FIG. 2 (bl and (C1), respectively.

The detection circuit 6 detects when a predetermined value or more in the projection distribution as shown in (bl and (C) in FIG. 2) extends over a predetermined range from the original image shown in FIG. A region corresponding to a continuous region (ΔX and Δy in the figure) is detected as a unique region.

If Fig. 3 is false, Fig. 2 (al) is stored in the image memory 1.
The identification circuit 8 identifies the pixels of the unique area detected from the original image such as 8.

The identification result is stored in memory 10.

Next, the identification circuit 9 connects the original image as shown in FIG. 2(a) stored in the image memory 1 to the pixels in the area shown in FIG. is 1~
15 pixels are identified and the identification results are stored in the memory 10.

The identification circuit 9 repeats the above process, and as a result, the memory 10
The pixels stored in are expanded sequentially as shown in FIG. 3 (bl), and finally become as shown in FIG. 3 [C1].

As can be seen by comparing Figure 2 ta+ and Figure 3 (C1).

The memory 10 stores only the identification result of the pixel corresponding to ■ out of the stored contents of the image memory 1. The corresponding pixel does not appear.

Thereafter, the selection circuit 11 extracts only the image data to be recognized, which is indicated by ■, out of the original images 1 and 2 of fa+ in FIG.

According to the above embodiment, pixels in an area corresponding to an area in which a predetermined value or more in the projection distribution is continuous over a predetermined range or more;
Only pixels with a predetermined density or higher that are connected to this are extracted from the original image, and are removed as frames or dirt.

In particular, stains with a density close to the density of the recognition target image but separated from the recognition target image are not extracted.

(g> As described in detail of the invention, 2) According to the present invention, the background or dirt is rarely extracted, and the recognition target image is accurately extracted.
4 image extraction circuits can be obtained.

[Brief explanation of the drawing]

FIG. 1 is a block diagram of an embodiment of the present invention, and FIGS. 2 and 3 are explanatory diagrams of the functions of each part in the embodiment. In the figure, 2 and 3 are projection circuits, 6 is a detection circuit, 8 and 9 are identification circuits, and 11 is a selection circuit. Figure 2 (α) (b2 V--] r -7 1111 (C) (C) r-'-'1 1 1

Claims (1)

[Scope of Claims] A projection circuit that obtains a projection distribution of pixels in a predetermined density range in original image data in which the density of each pixel arranged in a matrix is expressed by multi-value data. a detection circuit for detecting a singular region in the original image from a projection distribution f-1 obtained by the projection circuit; and a detection circuit for detecting a peculiar region in the original image from the projection distribution obtained by the detection circuit, and pixels in the predetermined density range sequentially connected to the peculiar region obtained from the original image by the detection circuit. An image extraction circuit comprising: an identification circuit that identifies a pixel; and a selection circuit that selects from the original image a pixel that corresponds to the pixel identified by the identification circuit.