JPS59135573A - Identifying system of photo region - Google Patents

Abstract

PURPOSE:To decide that the region of a discordance window of comparison is included in a photo region by observing the document pictures through a two-dimensional window of (mXn) picture element size together with a successive scan and comparing the variable density structure within the window with a basic edge pattern. CONSTITUTION:The variable density data of document pictures are stored in a buffer memory 1 by an amount equivalent to five lines and then delivered to a reference register 2 from upper three stages of the memory 1. The variable density data of a (3X3) picture element window are stored temporarily in the register 2 and then supplied to a pattern comparator 3 as well as to an arithmetic circuit 4. The comparator 3 compares the input picture element with the basic edge pattern and delivers a discordant output to a deciding circuit 6 if no coincidence is obtained from said comparison. Then the circuit 4 calculates the mean variable density value within the window from the (3X3) data and then compares this density value with the threshold value to apply an effective output to the circuit 6 when the mean variable density value is larger than the threshold value. The circuit 6 decides a region where (2X30) units of input signals continue as a photo region. Thus it is possible to identify the photo region in real time.

Description

DETAILED DESCRIPTION OF THE INVENTION [Technical Field] The present invention relates to a method for identifying photographic areas in a document image.

[Prior art]

Generally, the elements constituting a document image include sentences, figures, tables, graphs, and photographs. Among these, the photo &1 essentially require shading information, but the others can be handled as black and white binary information. Therefore, in order to faithfully and efficiently store, search, transmit, and record document images, it is necessary to distinguish between photographic areas and other areas in the document image, and to apply optimal processing to each area.

For example, in a facsimile, a recording mechanism can generally record only binary information, so it is necessary to convert document image data into a binary format. In this case, it is sufficient to simply binarize the areas other than the photo area in the document image using a certain predetermined threshold, but the photo area must be binarized using systematic dithering or the like.

As described above, when dealing with document images that include photographs, it is often necessary to identify photographic areas. However, conventional photo area identification methods, for example, separately perform binarization using a fixed threshold value and binarization using a systematic dithering method for the entire area of a document image, and then identify photo areas from the binarized data. This was extremely inefficient and difficult to implement in hardware.

〔the purpose〕

SUMMARY OF THE INVENTION An object of the present invention is to provide a new photographic area identification method that can be implemented efficiently using simple hardware and that can be easily processed in real time.

〔Example〕

In a document image, all parts of figures, tables, documents, etc. other than photographic areas can be considered to be line figures. A line figure is a collection of lines, and the two-dimensional shading structure at the edge portions (edges, j, ends) of the lines is limited. When a document image is observed through a window of m×n pixel size, the shading structure within the window at the edge portion exhibits a shading structure (basic edge pattern) unique to the edge portion. Furthermore, this basic edge pattern hardly occurs in photographic areas. Therefore, by sequentially scanning a document image with a window of m x n pixel size and comparing the shading structure within the window with the basic edge pattern, it can be determined whether the window area is the edge of a line or not.

The present invention focuses on this point, and observes a document image given as gradation data while sequentially scanning it with a two-dimensional window of m x n pixel size, and compares the gradation structure within the window with the basic edge pattern. The method is characterized in that the area of the window where the comparison does not match is determined to be included in the photographic area.

When implementing the present invention, it is preferable to take into account noise mixed in gray scale data of a document image. This noise includes the density and dirt of the background of the document image, and unnecessary shading information components such as characters on the back side.

To deal with such noise, it is sufficient to check the average gray value within the window and exclude windows for which the average gray value is less than a certain threshold value from the identification target.

Furthermore, there is a possibility that a shade structure other than the basic edge pattern may be observed inside the line. Although this possibility is generally small, in order to achieve highly accurate photographic area identification, it is necessary to accurately distinguish between the inside of a line and the photographic area. This purpose is achieved by determining that the window area is included in the photographic area only when there are 1Xk or more two-dimensionally consecutive windows that are inconsistent when compared with the basic edge pattern, as described below. can.

Hereinafter, the present invention will be specifically described in detail with reference to the drawings.

When a document image is observed through a 3 x 3 pixel size window, when the window comes to one of the positions shown in Figure 1, the shading structure within the window will match one of the basic edge patterns shown in Figure 2. Match. Here, in FIG. 1, a 3×3 mesh represents one window, and a shaded area represents a line. Furthermore, in FIG. 2, H means the condition that the gray value of the pixel must be greater than or equal to a certain threshold TH, and L means that the gray value of the pixel must be less than or equal to another certain threshold TTJ(T,) T1). means that the condition must be met. F means the condition that the gray value of the pixel may be arbitrary.

The document image is represented as 8-bit grayscale data, and the threshold value TH
Figure 3 shows the results of investigating the probability that 7Xk windows having a gradation structure other than turns (l is (in the sub-scanning direction, k is the number in the main-scanning direction), excluding windows whose average gray value is less than the threshold value TA=30.

(a), (b), and (C) in Figure 3 each show the probability distribution in the Japanese (Kanji/Na mixture) text area, and (a) is! =1, (b) is no 2, (C) is! =
This is the case where it is set to 3. (d) j (e) in Figure 3,
(f) shows the probability distribution in the English document area, (d) is /, , = 1, (e) & house A = 2, and (f) is the case where l = 3.'' From Figure 3. As is clear,
In a text area (1), it is virtually impossible for there to be more than 3 x (9) consecutive windows exhibiting a shading structure other than the basic edge pattern; We have confirmed through experiments that there is no problem even if it is determined to be a photographic area.

Next, an embodiment of the present invention will be explained based on the fourth factor.

1 is a buffer memory, which stores the density data (8
It has a five-stage shift register configuration so that it can store five lines of bits. The grayscale data is input from the top stage of the buffer memory 1 and is sequentially shifted to the lower stage. The grayscale data is output from the upper three stages of the buffer memory 1 to the reference register 2, and the grayscale data of 3×3 pixels is temporarily stored in the reference register 2. In other words, the document image is sequentially scanned in a 3×3 pixel size window, and the data of the shading structure within the window is obtained in the reference register 2. This 3
The grayscale data of ×3 pixels is input to a pattern comparison circuit 3 and an arithmetic circuit 4.

The pattern comparison circuit 3 compares the shading structure within the window represented by the input 3×3 pixel shading data with the basic edge pattern (mask pattern) shown in FIG. Outputs a mismatch signal. The arithmetic circuit 4 calculates the average gradation value TM within the window from the input 3×3 pixel gradation data and supplies it to the comparator 5.

This comparator 5 compares the average gray value TM with the aforementioned threshold value TA, and outputs a valid signal only when TM≧TA.

The above-mentioned mismatch signal and valid signal are input to the determination circuit 6. This judgment circuit 6 has a window (having a shading structure other than the basic edge pattern) into which the mismatch signal and the valid signal are input.
An area in which 2×30 or more consecutive windows (with an average gradation value equal to or higher than the threshold TA) are determined to be a photographic area, and a selection signal is output only during the period when the gradation data of the photographic area is output from the bottom stage of the buffer memory 1. .

The grayscale data outputted from the bottom stage of the buffer memory l is binarized by a fixed threshold value in a binarization circuit 7, and at the same time, it is binarized in a dither circuit 8 by a systematic dither method.

9 is a selection circuit which selects the output data of the dither circuit 8 during a period when a selection signal is given from the determination circuit 6 and outputs it as binary data; however, during a period when a selection signal is not given, the output data from the dither circuit 7 is selected. output data is selected and output as binarized data. In other words, in this embodiment, the gradation data of the photographic area is binarized using the systematic dithering method, and the gradation data of other areas is binarized using the fixed dither method.

〔effect〕

As described in detail above, the present invention has the effect of being able to identify photographic areas through real-time processing using a simple hardware configuration.

[Brief explanation of the drawings] Figure 1 is a diagram showing the positional relationship between the edge of the line and the window, Figure 2 is an explanatory diagram of the basic edge pattern, and Figure 3 is a diagram showing the positional relationship between the edge of the line and the window.
There are 7X windows with shading structures other than the basic edge pattern.
FIG. 4 is a schematic block diagram showing an embodiment of the present invention. 1...Buffer memory, 2...Reference register, 3.
...Pattern comparison circuit, 4. Arithmetic circuit, 5. Comparison circuit, 6. Judgment circuit. Procedural amendment (voluntary) Revenue stamp amount: 0 yen May 13, 1947, Director General of the Patent Office Kazu Sugi Failure 1, case indication Patent application No. 10373-1982 2, issued
Name Photographic area identification method 3, person making amendment Relationship to the case Applicant 4, agent ・151 Address 2-38-12 Yoyogi, Shibuya-ku, Tokyo
Kinkei Building No. 201 No. 5 Number of inventions increased by amendment None 6, Subject of amendment "Detailed explanation of the invention" column 7 of the specification, Contents of amendment "English document area" on page 6, lines 10-11 of the specification ” is corrected to “English sentence area”. that's all

Claims (3)

[Claims] (1) In a photo area identification method that identifies photo areas in a document image, a document image given as grayscale data is observed while being scanned through a two-dimensional window with a predetermined m x n al element size, and each window is The shading structure of the window is compared with a predetermined shading structure corresponding to the edge portion of the line, and if the comparison does not match, the area of the window is determined to be included in the photographic area in the document image. Photo area identification method. (2) A patent claim characterized in that only when a predetermined lx number or more of windows whose shading structures do not match are two-dimensionally continuous, the region of the window is determined to be included in the photographic region. The photographic area identification method described in item 1. (3) Claim 1, characterized in that only the region of the window having an average gray value equal to or higher than a predetermined threshold value is to be identified.
2. Photographic area identification method as described in Section 2 or Section 2.