JPH04105178A - Document picture processor - Google Patents

Abstract

PURPOSE:To reduce a storage capacity by separate-extracting a character area, a graphic area and an image data from document information fetched in the form of image information, converting the information included in the respective areas into code data suited to the information, after that, reconstituting and outputting them. CONSTITUTION:A picture input part 101, a picture memory 102, a preprocessing part 103, a feature extracting part 104, an area dividing part 105, a character area processing part 106, a graphic area processing part 107, an image area processing part 108, a reconstituting part 109 and a picture output part 110 are provided. Then, an inputted document picture is separated into a character area, a graphic area and an image area, moreover, a character is recognized in the character area, the character and a geometric graphic are recognized in the graphic area, they are converted into code data suited to the respective information and a document is stored. Thus, in the case of storing, a small storage capacity is satisfied.

Description

[Detailed Description of the Invention] [Industrial Application Field] The present invention separates and extracts character areas, figure areas, and image areas from document information captured in the form of image information,
The present invention relates to a document image processing device that converts information included in each area into code data suitable for the information, then reconfigures and outputs the code data.

[Prior Art] Generally, when document information written on paper is stored, an image captured by an image scanner or the like is treated as image information and is stored in an external storage device or the like. Also,
If only a character region is extracted, then the character region is subjected to character recognition, character coded, and saved, the user needs to manually specify the character region for the input document image.

[Problems to be Solved by the Invention] As described above, the type that saves in the form of image information requires a huge amount of storage capacity, and also cannot perform editing work such as rewriting part of the document. The type that specifies areas requires a human being to be present at all times and give instructions, which is time-consuming.

SUMMARY OF THE INVENTION The present invention aims to solve the above-mentioned problems by separating an inputted document image into a text area, a diagram area, and an image area without relying on humans. In this area, the object is to provide a device that recognizes characters and geometric figures, converts them into code data suitable for each type of information, and stores the document.

[Means for Solving the Problems] The document image processing device of the present invention includes an image input section that captures a document containing at least one of characters, diagrams, and image areas as image information, and A preprocessing unit that removes noise from image information and binarizes it; a feature extraction unit that extracts the features of text, diagrams, and image areas in the image information; and a feature extraction unit that extracts the aforementioned image based on the features extracted by this feature extraction unit. an area dividing unit that divides information into a character area, a diagram area, and an image area; a character area processing unit that recognizes characters in the character area; a diagram area processing unit that recognizes geometric figures and characters in the diagram area;
An image area processing unit that processes image information in an image area, a reconstruction unit that reconstructs data from each of these processing units, and an image output unit that outputs reconstructed document information. Features.

[Example] The present invention will be explained below based on the drawings.

FIG. 1 is a block diagram showing the configuration of a document image processing apparatus according to the present invention. An image input unit 101 inputs a document image as image information, and uses a scanner, a camera, or the like. When using an optical disk or the like on which images are stored in advance, the playback device is compatible with this. 1
02 is an image memory that temporarily stores captured image information. 103 is a preprocessing unit that removes noise contained in image information and binarizes it. To remove noise,
Remove isolated noise using a median filter or the like. If the input image is tilted, the tilt angle is corrected in this part to make later processing easier. Reference numeral 104 denotes a feature extraction unit that extracts feature quantities for separating character areas, diagram areas, and image areas. The feature extraction method will be described later.

Reference numeral 105 denotes an area dividing unit that divides the input image information into a character area, a diagram area, and an image area based on the feature amount extracted by the feature extracting unit 104. A character area processing unit 106 extracts a character string, cuts out a - character, and recognizes the cut out characters within the divided character areas.

Reference numeral 107 is a diagram area processing unit that performs recognition after extracting geometric figures and characters in the diagram area. Reference numeral 108 denotes an image area processing unit that processes a portion determined to be an image area as image data or compresses it.

Reference numeral 109 denotes a reconstruction unit that reconstructs the area converted into code data and the area of image data on the paper. Reference numeral 110 denotes an image output unit that outputs reconstructed document information, and specifically includes a printing device, a display device, and an external storage device.

Next, a method for extracting features of input image information will be described. First, the input image is grouped into m pixels in the vertical direction and n pixels in the horizontal direction. Then, the number of black points existing in each group (mXn pixels) is counted. This operation is performed on the entire input image.

Figure 2 shows that the input image information is divided into groups of mXn pixels, and the number of black pixels present in the groups is counted over the entire input image, and the density (number of black pixels) is plotted on the horizontal axis. FIG. 3 is a diagram showing a histogram with frequency (frequency) on the vertical axis. - For boat fishing, the chart area has a lot of white parts and the density is low. On the other hand, the image area has many black parts and has a high density.

The character area is located in between. As shown in the figure,
If the regions separated by appropriate thresholds (tl, t2, t3) are numbered 0.1.2.3 in descending order of density, then 0
can be divided into blank areas where nothing is written, 1 is a diagram area, 2 is a text area, and 3 is an image area. In this example, a document with a large character area was used as a sample, so the frequency of the portion corresponding to the character area is large.

FIG. 3 is a diagram showing an example in which an input document is labeled for each m x n pixel by the method shown in FIG. 2. The area labeled 1 is the diagram area, 2
corresponds to the character area, and 3 corresponds to the image area. In this way, region division is performed by grouping regions with the same label.

FIG. 4 is a block diagram showing details of the character area processing section 106. 41 is a character string extraction unit that extracts a character string within a character area; 42 is a character extraction unit that extracts one character from the extracted character string; and 43 is a character recognition dictionary 4 that extracts the extracted character.
4, it is a character recognition unit that performs recognition. Characters that could not be recognized here are sent as image data to the next word matching section 45. The word matching unit checks whether the recognized characters have meaning as words or not by referring to the word dictionary 46 and if there are meaningless words due to erroneous recognition of characters, check the words that can be corrected. Convert to the correct word. Characters that cannot be recognized by the character recognition unit but can be determined by referring to the word dictionary are determined in this part. Characters that cannot be determined by the word matching section are left as images. 47 is a character encoding unit that encodes recognized characters. 48 is a true format addition section that adds information necessary for reconstructing the character string converted into code data in the above section on paper. for example,
As shown in Figure 5, when the origin is the upper left of the page, the X axis is vertical and the Y axis is horizontal, (xLy
Suppose that a character string exists in an area surrounded by (x2, y2).

At this time, the page format showing this area is, for example, as shown in FIG.
). In this example, ++ Char”
However, this area is a character area, and the position of the character area is within a rectangle surrounded by (xl, yl) and (x2, y2), and the character type is Mincho typeface. The font size is 10 points, and the recognized character string is (@@
・Old・River・・・・Represents the content indicated by @@).

Based on this information, the information in the character area is output to a printing device or display device, or stored in an external storage device.

FIG. 7 is a block diagram showing details of the chart area processing section 107. 71 is a geometric figure extraction unit that extracts geometric figures in the chart area, and the extraction method is H.

Ugh conversion, connected component extraction of black pixels, etc. are used.

72 is a character extraction unit that extracts characters included in the diagram area;
Reference numeral 73 denotes a character recognition unit that refers to the character recognition dictionary 74 to recognize the extracted characters. Characters that could not be recognized here are sent as image data to the next word matching section 75.

The word matching section refers to the word dictionary 76 to determine whether the recognized characters have meaning as words, and if there are meaningless words due to misrecognition of characters, the words that can be corrected are corrected. Convert to word. Characters that cannot be recognized by the character recognition unit but can be determined by referring to the word dictionary are determined in this part. For characters that could not be confirmed by the word matching section,
Leave it as an image. 77 is a character encoding unit that encodes recognized characters. 7
Reference numeral 8 denotes a true format addition section that adds information necessary to reconstruct the character string converted into code data in the above section on paper. For example, as shown in Figure 5, when the origin is at the top left of the paper, the X axis is vertically, and the Y axis is horizontally, the straight line connecting (x3, y3) and (x4, y4) is Suppose it exists. At this time, the page format showing this straight line is
For example, it becomes as shown in FIG. 6(b). In this example,
"Line" indicates that this figure is a straight line, that the position of the straight line is in the area connecting (x'3, y3) and (x4, y4), that the type of line is a solid line, and that the line Width is 0
．． Each indicates that it is 5 mm. In FIG. 7, portions 72 to 78 may be shared with those included in the character area processing section 106.

The following can be considered as application examples of the present invention. In the electronic filing system, data can be compressed by dividing the Kugata image into regions and encoding it.
Storage capacity can be reduced. By combining it with desktop publishing, it can be used to rewrite text and graphics in an input image to create another document.

When performing machine translation, it is possible to automate document input that previously had to be done manually using a keyboard. Conventionally, when copying machines repeatedly reproduce image information, characters and figures become blurred due to noise and become unreadable, but this time, through this document processing device,
By coding the parts that can be coded,
No matter how many times the coded part is copied, a clear image can always be obtained. Furthermore, for the same reason, when used to process facsimile input images, the images become clearer and the transmission capacity is reduced.

[Effects of the Invention] As described above, when the document image processing device of the present invention is used, characters and figures that were conventionally handled as image data are recognized and converted into code data suitable for them, so it is easier to save them. In this case, less storage capacity is required. Also, since it has been converted to code data, editing work such as changing or reusing a portion of characters, figures, etc. can be performed.

Furthermore, since text, diagrams, and image areas are automatically separated and extracted, not only is it possible to save processing time, but by setting the program in advance, it is possible to erase logo marks that exist in the margins, etc. You can also perform trick plays such as saving only text or shapes.

[Brief explanation of drawings]

FIG. 1 is a block diagram showing the configuration of the document image processing device of the present invention, FIG. 2 is a diagram showing the distribution of black pixels in each region, FIG. 3 is a diagram when an input image is divided into regions, and FIG. 4 is a block diagram of the character area processing unit of the present invention, FIG. 5 is a diagram showing an example of an input image, FIG. 6 is a diagram expressing FIG. 5 in true format, and FIG. 7 is a diagram showing an example of an input image.
The figure is a block diagram of the chart area processing section of the present invention. 101... Image input section, 102... Image memory, 1
03... Preprocessing unit, 104... Feature extraction unit, 105
...Area dividing section, 106...Character area processing section, 10
7... Chart area processing unit, 108... Image area processing unit, 109... Reconstruction unit, 110... Image output unit, 41... Character string extraction unit, 42, 72... Character Extraction unit, 43.73 Character recognition unit, 44.74 Character recognition dictionary, 45.75 Word matching unit, 46.7
6... Word dictionary, 47.77... Character encoding section,
48, 78...Page format addition section, 71...Geometric figure extraction section Applicant: Seiko Epson Co., Ltd. Agent, Patent Attorney Dozu Flood Damage Department 3 (1 other person) t2 Density Figure 3 Figure 6 (b)

Claims (1)

[Claims] an image input unit that captures a document containing at least one of text, diagrams, and image areas as image information; a preprocessing unit that removes noise from the image information captured by the image input unit and converts it into binary; and the image information. a feature extraction unit that extracts features of the text, diagram, and image regions; and an area division unit that divides the image information into text, diagram, and image regions based on the features extracted by the feature extraction unit; a character area processing unit that recognizes characters in the character area; a diagram area processing unit that recognizes geometric figures and characters in the diagram area; an image area processing unit that processes image information in the image area; 1. A document image processing device comprising: a reconstruction section that reconstructs data from three processing sections; and an image output section that outputs reconstructed document information.