JPH07192083A - Document picture layout analysis device - Google Patents

Abstract

PURPOSE:To perform highly accurate layout analysis to a document in which the characters of different point numbers coexist. CONSTITUTION:The connection components of document pictures inputted from a document picture input means are extracted by the extraction means 11 of the connection components and classified corresponding to size information by the classification means 12 of the connection components and picture planes provided with only the connection components.belonging to a class obtained by performing the classification are respectively generated by a partial picture generation means 13. The layout analysis is performed to the respective divided picture planes by layout analysis means 14 and 15, layout information extracted from the respective picture planes is synthesized in a layout information synthesis means 16 and a layout analyzed result over the entire document pictures is obtained. When discrepancy is generated at the time of synthesizing the layout information, the layout information of the plane provided with a lot of the connection components and provided with the connection components in a size appropriate for the characters is preferentially synthesized and the final layout analyzed result is obtained.

Description

Detailed Description of the Invention

[0001]

BACKGROUND OF THE INVENTION 1. Field of the Invention The present invention relates to a document image layout analysis apparatus.

[0002]

2. Description of the Related Art In a document image recognition apparatus, layout analysis processing is performed on a digital image obtained by inputting a document by image scanning or the like, and character recognition processing is performed on each extracted character image. The process flow of converting to a character code is taken. Here, the layout analysis means a process of extracting a character line from an input image and cutting out a character.

As a document relating to the layout analysis methods that have been proposed so far, "Document recognition system using highly accurate character segmentation processing based on character structure information" (Sun et al., Information Processing Society of Japan, Vol.33, No. .99,
1992). This is a method for Japanese documents, and its outline is explained below. In this method, the input document image is subjected to binarization and inclination correction processing, and then subjected to layout analysis. Region division is performed, only the character region is extracted, and the black pixel connected component is extracted by performing a labeling process on the character region. The obtained block of black pixels is called a candidate figure. There are many separated characters such as “north” and “c” (characters composed of multiple connected components) and contact characters that make contact to form one connected component. A process to generate a figure is required. The most frequently used size of the corrected figure is the average size of the characters. The candidate figure is projected on the horizontal axis, and examples are extracted from the distribution. Integrate candidate shapes in columns to deal with separators. The candidate figure with a size close to the average size is preferentially cut out, the remaining figures are forcibly integrated for separated characters, separated for contact characters and extracted for punctuation, and the final character cutout result To get

Another layout analysis method is
As an example for English documents, TSUJIMOTO
By "Major Components of
aComplete Text Reading S
system ”(TUJIMOTO et al., Proceedi
ngs of the IEEE, Vol. 80, N
o. 7, 1992). The document image is expressed by run length, (1) the connected component of the black pixel is extracted, and the adjacent connected components at a relatively short distance of about 1 mm are integrated and segmented. (2) Classify each segment into text lines, figures, pictures, etc. (3)
Integrate adjacent segments of text lines into segments. This extracts a block of text. Here, the words can be extracted by the processes of (1) and (2), and the text lines can be extracted by the process of (3).

According to these methods, when character lines having different points are present close to each other in the same document, they cannot be correctly extracted. In the layout analysis process, the typesetting rule of the document that the space between columns is larger than the space between lines, the space between lines is larger than the space between characters, and the space between characters is larger than the space inside characters is established in the input document image. The layout analysis of the document image is performed on the assumption that the above is done. At this time, for example, when they exist closer to each other than the character intervals having different points as shown in FIG. 10, the line intervals of the text lines having the smaller points become smaller than the character intervals of the characters having the larger points, and the points having a larger number of adjacent points become larger. Characters are merged into a text line with a small number of points, making it impossible to extract the correct line and character.

As described above, when characters having different points are mixed in the same document, layout analysis is difficult.

[0007]

In the conventional layout analysis process, the space between columns is larger than the space between lines, the space between lines is larger than the space between characters, and the space between characters is larger than the space within characters. Only when the typesetting rule of is valid for the input document image,
The layout analysis of the document image worked properly. In a document in which characters having different points are mixed, such a rule does not always hold.

An object of the present invention is to realize highly accurate layout analysis performance for a document in which characters having different points are mixed in the same document, which is difficult with the conventional method.

[0009]

The present invention is obtained by classifying document image input means, means for extracting connected components from the document image, and means for classifying the connected components according to size information. Means for respectively generating partial images including only connected components belonging to the class, and for each partial image,
It is characterized by including a layout analysis means for performing layout analysis and extracting layout information, and a means for synthesizing the layout information.

[0010]

The connected components are classified according to the size information, and the partial images including only the connected components belonging to the obtained class are generated, so that the characters having different points are distributed to the partial images. Layout analysis is performed on each partial image by the conventional method. Further, the layout analysis obtained from the result of the layout analysis of each partial image is performed. Further, the final layout analysis result is obtained by synthesizing the layout information obtained from the layout analysis result of each partial image. When there is a contradiction between layout information when synthesizing layout information, priority is given to the layout analysis result of the class that includes the connected components that have a large number of connected components in each partial image and have a character-like size. To synthesize.

[0011]

DETAILED DESCRIPTION OF THE PREFERRED EMBODIMENTS Embodiments of the present invention will be described below with reference to the drawings.

FIG. 1 is a block diagram showing the same embodiment.
The document image input by the image input processing is binarized and converted into a scanning signal of the image. FIG. 3 shows an example of an image obtained by scanning an input document image and then binarizing it. This input document contains the characters corresponding to the main text that occupies most of the title document, as well as four components with different sizes: large characters corresponding to the title part, small ruby characters, and granular noise. There is something.

Each black pixel connected component is extracted by the connected component extracting means 11 to obtain circumscribed rectangle information 101 of the connected component. Here, the black pixels correspond to elements such as characters and figures that form a document, and the white pixels correspond to the background portion. Next, if the two circumscribing rectangles have an inclusive relation, or if the degree of overlap of the circumscribing rectangles is greater than a certain degree, then both rectangles are integrated. Here, the degree of overlap is represented by, for example, the ratio of the area of the overlapping portion to the total area of two overlapping rectangles. FIG. 4 shows a result of performing integration of rectangles having an inclusion relation and integration using a degree of overlap with respect to a circumscribed rectangle of connected components.

The connected component classifying unit 12 classifies each connected component into a class according to the size of the circumscribed rectangle of the connected component extracted and integrated by the connected component extracting unit 11, and classifies the connected component into class information. Output. Here, a class is a set of connected components having a similar size, for example, a class C ₃ corresponding to a heading character and a class C corresponding to a main character.
_2. There are four possible types: class C ₁ corresponding to ruby and lettering, making, punctuation, and class C ₀ corresponding to small noise. The height, width, area, contour length, and the like of the circumscribed rectangle can be considered as the criteria for classification, but here, as an example, classification into classes is performed using the area of the rectangle. Specifically, the area of each rectangle is calculated and, for example, K-m
Each rectangle is classified into classes C ₀ , C ₁ , C ₂ , C ₃ , ... By clustering using the eans method (for example, K = 4).

On the basis of this, the partial image generating means 13 allocates the connected component belonging to each class to each partial image, and generates the partial images 103 to 104 for each class. For example, a partial image that includes only connected components belonging to the class C _i is P _i (i = 1, 2, 3). FIG. 5 (a)-
(D) is the result of assigning a connected component to each partial image according to the classification information on the class of the circumscribed rectangle. In this example, the partial image P ₀ is the noise, the partial image P ₁ is the letter characters including separation in ruby and characters remained separate without being integrated process during rectangular integrated portion, the partial image P ₂ is present The character of the minute portion, the partial image P ₃ corresponds to the character of the title portion.

The converted partial images are subjected to layout analysis processing by the corresponding layout analysis means 14 to 15.
In the partial image layout analysis means here, since rectangles having substantially the same rectangular area are already allocated to the respective partial images, the method of Son et al., Tujimot described in the prior art.
It is possible to use the method of O et al. Here, as an example, the method proposed by Tsuji in 1985, entitled “Structural Analysis of Page Image Based on Split Detection Method”, was proposed in the Technical Report of Electronic Communication Studies, Pattern Recognition and Learning, PRL85-17, for partial images of each class. Commonly applied. In the layout analysis method using this method, the projection pattern is obtained in each of the partial images in the horizontal and vertical directions. , Estimate character pitch from line width. Using the estimated character pitch, characters are cut out from the character line at approximately equal intervals in consideration of some variation. By using such a conventional technique, the interspersed characters are placed in a positional relationship,
Characters and lines can be extracted in consideration of the periodicity of the character arrangement. FIGS. 6A to 6D are the results of extracting character lines and characters from the partial images of each class when the same input image example is input.

Layout information 105 to 10 for each partial image
The layout information synthesizing means 6 synthesizes the layout information 6 to output the final layout analysis result of all the documents.
An embodiment relating to the layout information synthesizing means 16 will be described with reference to FIG.

First, the connected component number histogram generating means 2
A connected component number histogram 200 which is a histogram of the number of rectangles belonging to each class is obtained from 0, and the number of connected components included in each class is stored. FIG. 7 is a histogram of the number of connected components when the input image example of FIG. 3 is input.

In the layout information selecting means 22, when the layout information is synthesized, the layout information of each partial image is simply weighted. As a result, when one connected component causes a contradiction in a plurality of layout information, priority is given. A priority request signal 202 is issued to the degree determining means 21.

The layout information selecting means 22 synthesizes each layout information. At this time, if an inconsistent portion of the layout information is found, a priority request signal 202 is generated and sent to the priority determining means 21. Next, according to the priority signal 201 obtained from the priority determining means 21, the layout information of the class having a high priority is selected.
This is performed for all inconsistencies regarding the layout information, and the layout information synthesis result 107 is output.

The priority determining means 21 uses the priority request signal 20.
When 2 is received, the connected component number histogram 200 is referred to, each frequency in a plurality of classes corresponding to the portion in which the layout information is inconsistent is compared, and the class having the maximum frequency is subjected to layout information synthesis. Priority is given. Also, at this time, when the size of the circumscribed rectangle of the connected component included in the class is small and most of the connected components included in the partial image are regarded as noise, or when the size of the circumscribed rectangle is large and the connected component of the connected image included in the partial image is large. If many are regarded as figures, tables, etc., no priority is given.
The resulting priority signal 201 is returned to the layout information selection means 22.

FIG. 8 shows the result of simply superposing the layout information of the partial images of each class. In the figure, there is a contradiction regarding the layout information in the horizontal stroke of the character “line”, the right side of the “ka”, and each connected component of the “Noh”.
Referring to the frequency of the number of characters belonging to the partial images P1 and P2 involved in this contradiction, the frequency of the number of characters belonging to the partial image P2 is high. In this part, the layout analysis result of the partial image P2 is given priority, Priority is given to layout information composition. Further, since the connected components belonging to the partial image P0 are considered to have a small rectangular size and contain a large amount of noise components, priority is not given. Therefore, when the layout information is combined, connected components other than those that overlap with the layout information of other partial images are deleted. FIG. 9 is a result of synthesizing layout information, and a correct layout analysis result is obtained regardless of a document in which characters having different points are mixed.

With the above method, the layout analysis result for the entire document can be obtained without being restricted by the size of the character.

[0024]

As shown in FIG. 10, for example, in a document in which lines composed of characters having different points are close to each other, partial images are generated for each size of connected components, and layout analysis processing is performed for each partial image of each class. The correct analysis result can be obtained by performing.

[Brief description of drawings]

FIG. 1 is a block diagram showing a configuration of a layout analysis method according to an embodiment of the present invention.

FIG. 2 is a block diagram according to an embodiment of layout information synthesizing means in FIG.

FIG. 3 is an example of a document image input.

FIG. 4 is a circumscribed rectangle extraction result of the connected components extracted by the connected component extracting means when the document image of FIG. 2 is input.

FIG. 5 is a result of generating a partial image by dividing connected components classified into each class by a partial image generating unit into each partial image with respect to the same input image. (A) Partial image P
₀ , (b) partial image P ₁ , (c) partial image P ₂ , (d)
Partial image P ₃ .

FIG. 6 is a layout analysis result of each partial image with respect to the same input image; (a) a layout analysis result of a partial image P ₀ ; (b) a layout analysis result of a partial image P ₁ ;
Layout analysis result of partial image P ₂ , (d) partial image P
Layout analysis result of ₃ .

FIG. 7 is a connected component histogram as a result of counting the number of connected components classified into classes by a connected component classification unit when the document image of FIG. 2 is input.

FIG. 8 is a result of superimposing layout analysis result information of partial images of each class on the same input image by a layout information synthesizing unit.

FIG. 9 is a result of integrating the layout information of the same input image by laying out the layout information superimposing result of FIG. 7 by giving priority to the layout information of the class having a high priority.

FIG. 10 is a conceptual diagram regarding an effect when the present technique is applied to an example of a document image including characters having different points in the same line. (A) An original image, (b) a partial image of a class in which a character with a large number of points is extracted, and (c) a partial image of a class in which a character with a small number of points is extracted.

[Explanation of symbols]

 11 Connected Component Extracting Means 12 Connected Component Classifying Means 13 Partial Image Generating Means 14 to 15 Layout Analyzing Means 16 Layout Information Composing Means 20 Connected Component Number Histogram Generating Means 21 Priority Determining Means 22 Layout Information Selecting Means 101 Circumscribing Rectangle Information 102 Classification information of connected components into classes 103 to 104 Image planes 105 to 106 Layout information of image planes 107 Layout information synthesis result

Claims (2)

[Claims] 1. A document image is input as a binary image signal,
Connected component extraction means for extracting connected components, connected component classification means for classifying the connected components according to size information, and only connected components belonging to a class obtained by performing classification according to size information Partial image generation means for respectively generating and storing the partial images including the layout images, layout analysis means for performing layout analysis on each of the partial images and extracting layout information, and layout information composition for composing each layout information. And a document image layout analysis device comprising means. 2. The document image layout analyzing apparatus according to claim 1, wherein the layout information synthesizing unit measures a frequency distribution of connected components for each class from layout information output by each layout analyzing unit, When a contradiction occurs between the connected component number histogram generation means stored as a number histogram and the layout information output by each layout analysis means, the priority when performing layout information synthesis with reference to the connected component number histogram is set. Priority determining means for determining, and laying out the layout information by prioritizing the layout information according to the priority,
A document image layout analysis device comprising: layout information selecting means for outputting final layout information.