JPS62123574A - Picture processing system - Google Patents

Abstract

PURPOSE:To correctly separate a character area and a photograph area by deciding whether the picture is the picture element to constitute a character or the picture element to constitute the photograph for respective picture elements, generating a binary plane and executing the filtering processing to the binary plane. CONSTITUTION:A picture information outputted from a picture reading device 11 is accommodated and given to a characater/photograph picture element deciding part 12. A deciding part 12 whether the picture information is a character picture element or a photograph picture element for respective picture elements, and in case of the character picture element 1, is outputted and in case of the photograph picture element, zero is outputted. The deciding result is accommodated to a binary plane memory 13. For the contents of the memory 13, the digital filtering processing is executed. As the result, when the mesh value in a filtering work memory 16 does not change, the contents of the memory 16 are successively read, given to a processing changing-over part 21, and synchronizing with this, the picture information is successively read from the memory 20. The changing-over part 21, when the contents of the memory 16 are one, communicates the picture information of the memory 20 to a binary processing part 22 and in case of zero, communicates it to a multi-value processing part 23.

Description

[Detailed Description of the Invention] [Technical Field] The present invention relates to an image processing method, and in particular, converts a read image including a binary image such as a character and a multi-gradation image such as a photograph into a binary image area or a multi-gradation image area. This paper relates to an image processing method for extracting .

[Prior art]

2. Description of the Related Art Facsimile machines and image processors often handle documents that include binary images such as characters (hereinafter collectively referred to as characters) and multilayer images such as photographs (hereinafter generally referred to as photographs). In such a case, if the read image can be divided into text areas and photo areas, image processing, data compression, etc. optimal for each area can be performed, leading to improved performance of facsimile machines, image processors, etc. Conventionally, there is a method of binarizing a read image at a certain level and then dividing it into text areas and photographic areas based on the distribution of run lengths, but accurate division cannot be expected.

〔the purpose〕

An object of the present invention is to accurately extract a character area or a photographic area from a read image of a document or the like containing one character and a photograph in a facsimile machine, an image processor, or the like.

〔composition〕

The present invention determines for each pixel in an image to be processed whether the pixel is a pixel that constitutes a character (text pixel) or a pixel that constitutes a photograph (photo pixel).
, a photo pixel is characterized in that, for example, a binary plane of "0" is generated, and a filtering process is performed on the two planes to extract a text area or a photo area.

An embodiment of the present invention will be described below with reference to the drawings.8 Fig. 1 is a block diagram of an embodiment of the present invention. The image reading device 11 sequentially reads the document 10 in a one-dimensional scanning format, converts it into multi-value data by analog-to-digital conversion, etc., and outputs image information corresponding to the density of each pixel. The image information in pixel units output from the image reading device 11 is sequentially stored in the 9-image memory 20. Here, it is assumed that the manuscript 10 contains a mixture of text and photographs as shown in FIG. A digital image corresponding to the document 10 is stored in the image memory 20 . In the following, it is assumed that the character area on the document]-〇 is extracted.

The image information output from the image reading device 11 is stored in the image memory 20 and simultaneously provided to the text/photo pixel determination section 12. The text/photo pixel determining unit 12 determines for each pixel whether the image information is a text pixel or a photo pixel, and if it is a text pixel, it is II I II, and if it is a photo pixel, it is 1/O.
Output r*. In principle, the determination as to whether a pixel is a text pixel or a photographic pixel is based on the density difference between the pixel of interest and surrounding pixels, but the details are given in Japanese Patent Application No. 67399/1983.
The explanation in 2\ is omitted. Text/photo pixel determination unit] The determination result in step 2 is stored in the binary plain memory 13.

Each storage location of the binary plain memory 13 is stored in the image memory 20.
If the pixel at the corresponding storage location is a character pixel, "Pa1" is stored, and if it is a photo pixel, "0" is stored.In FIG. Although character/photo pixels are determined directly from the information, the determination may be made after the image information is stored in the image memory 20.

Text/photo pixel determination is performed for each pixel above.

By assigning character pixels as "1" and photograph pixels as It OPI, the original image in FIG. 2 becomes as shown in FIG. 3. That is, when the photograph is removed, only the character image remains.In the present invention, a digital filtering process is applied to this image to finally extract a character area as shown in FIG.

The digital filtering process will be explained below.

First, the binary plain memory 13 is divided into MXN meshes, and the binary plain memory 13 is sequentially divided into MXN meshes.
The mesh size register 14 cuts out the contents of 3 into the mesh size register 14. The mesh size register 14 cuts out the contents of 3
It consists of the number of bits corresponding to The mesh determination unit 15 counts the number of uVt<black dots) present in the mesh (mesh (my n)) cut out in the mesh size register 14, and determines that the number is a predetermined value (for example, If one mesh is 4×4 bits, then 4) If larger, mesh (m, n)←1 otherwise.

mesh (m+n)←O is stored in the filtering work memory 16.

This is performed for all meshes sm,n (where 1≦m≦M, 1≦n≦N) that are cut out from the binary plain memory 13 to the mesh size register 14.

Through this processing, the picture constituent pixels are tentatively determined to be text pixels in text/photo pixel determination, and the binary plain memory 13
The black point stored as II I II is removed. Filtering work memory 16 is binary plain memory 13
It has the same bit configuration as . Therefore, although the binary plain memory 13 and the filtering work memory 16 are separated in FIG. 1 for convenience of explanation, this is not necessary.

Next, one mesh (referred to as a target mesh) and a plurality of meshes adjacent thereto are read out from the filtering work memory 16 to the mesh register 17. The mesh register 17 consists of, for example, a 3×3 mesh. The calculation/judgment unit 18 performs the following process A or B on the target mesh in the mesh register 17 using adjacent meshes, and stores the result in the filtering work memory 16 again. This is done for m from 1 to M and n from 1 to N.

Processing A: D←5Xmesh (m, n) -+++es
h(m-1,n)-+aash(m,n-1)-m
esh(m, n+1)-mesh(m+1.n)
-wsh (m+1. n+1) if D≧5 mesh (my n)←O if D≦-2 mesh (m, n)←1 Do nothing otherwise (-1≦D≦4).

Processing B: D4-5Xmesh(m, n)-me
sh(m-L n-1)-mesn-1)-, n)-
mssh(m, n-1)-mn-1)-, n+1)
mesh (m+1. n) if D≧5 mesh (my n)←O if D≧-2 mesh (m, n) 4-1 Do nothing otherwise (-1≦D≦4).

Processes A and B are performed alternately, for example, twice each, and the results are shown in FIG.

Next, the following process C is performed in the same manner through a loop of the filtering work memory 16, the mesh register 17, and the calculation/judgment section 18.

Processing C:D←4:Xsesh(m, n) -wsh(
m-1,n)-&1sh(men-1)-mesh
(m, n+1)-mesh (m+ 1, n) if D≧4 sash (m, n) ←Q if D≦-2 w3sh (m, n)←1 What else (-1≦D≦3) Neither.

In process C, for example, m initially ranges from 1 to M.

This process C is performed for n from 1 to N, then for m from M to 1, and for n from N to 1.
By repeating the process until there is no change in the value of h, the image shown in FIG. 5 is obtained.

The control unit 19 controls the reading from the filtering work memory 16, changes the weighting coefficient of the calculation in the calculation determination unit 18, and controls the repetition of processing.

When there is no change in the mesh value in the filtering work memory 16 (filtering is complete), the calculation determination unit 1
8 notifies the control unit 19 of this fact.

In response to this, the control section 19 sequentially reads out the contents of the filtering work memory 16 and provides them to the processing switching section 21, and in synchronization with this, sequentially reads out image information from the image memory 20. When the content of the filtering work memory 16 given through the control unit 19 is '1'' (character area), the processing switching unit 21 transmits the image information of the image memory 20 to the binary processing unit 22, and in the case of 140 II ( photo area) is transmitted to the multi-value processing section 23. The binary processing section 22 is suitable for processing characters, etc., and the multi-value processing section 23 is suitable for processing photographs, etc. The processing result of the multivalue processing section 23 is outputted via the disassembly section 24.

The configuration shown in Figure 1 can also be realized by a so-called microcomputer architecture, in which case the memory 1
2, 16.20 uses RAM.

〔effect〕

As described above, according to the present invention, it is possible to accurately extract and classify text areas or photo areas from a read image of a document or the like containing text and photos.

The performance of facsimile machines and image processors will be significantly improved.

[Brief explanation of drawings]

FIG. 1 is a configuration diagram of an embodiment of the present invention, and FIGS. 2 to 5 are image transition diagrams for explaining the operation of FIG. 1. 11... Image reading device, 12... Text/photo pixel determination section, 13... Binary plain memory. l4... Mesh size register, 15... Mesh judgment unit, 16... Filtering work memory, 17... Mesh register, 18... Calculation judgment unit, 19... Control unit, 20. ...Image memory, 21...Processing switching unit, 22...Binary processing unit, 2
3...Multi-value processing section, 24...Disabling section. Figure 2 Figure 3 Procedure Uichi 4 - (See 3 items) November 20, 1985 Director General of the Japan Patent Office Black 1) Akio Tono 1, Indication of the case 1985 Patent Application No. 263579 2. Name of the invention Image processing method 3. Relationship with the person making the amendment Applicant address 3-6, Nakamagome 1-chome, Ota-ku, Tokyo Name
(674) Rico Co., Ltd. - Representative
Hama 1) Hiro 4, Attorney 5, Number of inventions increased in the amendment None 7, Contents of the amendment The statement of the claims is amended as per the attached sheet. 8. List of attached documents A document stating the scope of patent claims 1 copy Claims (1) In a method for processing images that include a mixture of binary images such as characters and multi-tone images such as photographs, the subject of processing For each pixel of the image, it is determined whether the pixel is a binary image constituent pixel or a multi-tone image constituent pixel, and tri is applied to the binary image constituent pixel.
l+ (or 110 I+). “0” (or 111 I+) for multi-tone image constituent pixels
An image processing method characterized in that a binary plane is generated in which a binary plane is assigned, and a binary image area or a multilayer tone image area is extracted by performing digital filtering processing on the binary plane. (2) The digital filtering process divides the binary plane into a plurality of meshes, and for each mesh, when the existence state of "1" in the mesh satisfies a predetermined condition, the mesh is divided into 1''', Otherwise “0”
After that, for each mesh, refer to the meshes adjacent to it, and when a predetermined number or more of the adjacent meshes are "1", "1'", all the adjacent meshes are LL.
The first claim characterized in that when OI+ is 11 Q 11, nothing else is done.
Image processing method described in section.

Claims (2)

[Claims] (1) In a method for processing images in which binary images such as characters and multi-gradation images such as photographs coexist, the image to be processed is checked for each pixel to determine whether the pixel is a binary image constituent pixel or a multi-gradation image. It is determined whether it is a tone image constituent pixel, and “1” is set for a binary image constituent pixel.
(or "0") and "0" (or "1") to the pixels constituting the multi-gradation image, a binary plane is generated, and a digital filtering process is applied to the binary plane. An image processing method characterized by extracting a value image area or a multilayer tone image area. (2) The digital filtering process divides the binary plane into a plurality of meshes, and for each mesh, when the existence state of "1" in the mesh satisfies a predetermined condition, the mesh is "1", Otherwise “0”
After that, for each mesh, refer to the meshes adjacent to it, and set it to "1" when more than a predetermined number of adjacent meshes are "1", "0" when all the adjacent meshes are "0", and other meshes. 2. The image processing method according to claim 1, wherein the process of doing nothing is repeated until there is no change in the mesh value.