JPH08202824A - Document picture recognition device - Google Patents

Abstract

PURPOSE: To automatically edit a document and to reconstruct it by describing an indication mark on the document and designating an editing operation. CONSTITUTION: A color extraction part 1 dividing input picture data having a character and the indication mark into a character binary picture and an indication mark binary picture by the color, an indication mark segment part 2 taking out indication mark position/size information from the indication mark binary picture, an indication mark recognition part 3 recognizing indication information for the editing processing based on the indication mark binary picture and indication mark position/size information, a character segment part 4 taking out character position/size information from the character binary picture, a character recognition part 5 recognizing character information based on the character binary picture and character position/size information and a document constitution part 6 executing the editing processing corresponding to indication information based on indication mark position/size information, indication information and character position/size information and character information and reconstructing and outputting the document are provided.

Description

Detailed Description of the Invention

[0001]

BACKGROUND OF THE INVENTION 1. Field of the Invention The present invention relates to a document image recognition apparatus for reading characters from document image data of general documents such as newspapers and magazines and converting them into text data.

[0002]

2. Description of the Related Art In a conventional document image recognizing device, processing such as character / figure separation is performed on an input document image, and character patterns are cut out and recognized one by one in an area determined to be a character. According to such a conventional technique, the input document image is simply converted into text data as it is, and if the user wants to correct or edit a part of the text and convert it to data, instead of converting the document as it is. After performing the recognition process once, the text was manually corrected.

[0003]

However, according to the conventional technique, it is necessary to manually intervene in the correction process after the document image recognition, and the originals cannot be batch processed continuously, resulting in high efficiency. There was a problem of not having.

[0004]

SUMMARY OF THE INVENTION According to the present invention, input image data consisting of a character and an instruction symbol having a color different from the character and used for editing the character is converted into a character binary image by the color. Based on the color extraction unit that separates the indicator symbol binary image, the indicator symbol cutting unit that extracts the indicator symbol position / size information from the indicator symbol binary image, and the indicator symbol binary image and the indicator symbol position / size information. Based on the character binary image and the character position / size information, the character recognition unit that recognizes the instruction information for the editing process, the character cutout unit that extracts the character position / size information from the character binary image Based on the character recognition unit that recognizes the character information, the instruction symbol position / size information, the instruction information, the character position / size information, and the character information, the editing process corresponding to the instruction information is performed to reconstruct the document. Document structure part to be output as Characterized in that it has.

[0005]

The color extracting section divides the input image data into the reference symbol binary image and the character binary image by dividing the input image data by the color. The instruction symbol cutout unit receives the instruction symbol binary image and outputs the instruction symbol position / size information. The instruction symbol recognition unit receives the instruction symbol binary image and the instruction symbol position / size information, recognizes the instruction information, and outputs it to the document configuration unit.

The character cutout unit receives a character binary image as an input,
Outputs character position and size information. The character recognition unit receives the character binary image and the character position / size information, recognizes the character, and outputs it to the document configuration unit. The document composition unit inputs the instruction symbol position / size information, instruction information, character position / size information and character information, and reconstructs the document by editing the character information according to the editing operation specified by the instruction symbol. And output.

[0007]

Embodiments of the present invention will be described below with reference to the drawings. FIG. 1 is a block diagram of an embodiment and shows an outline of the configuration of the embodiment of the present invention. In the figure, reference numeral 1 is a color extraction unit, which inputs multi-valued image data of each of red, green and blue components of a color document image obtained by scanning a color document using a color scanner or the like (not shown). And The character color and the indicator color are different and can be distinguished. The character color component and the indicator color component are extracted from the input image data to obtain respective binary images, and the character binary image and the indicator binary value are obtained. Output as images respectively.

The above-mentioned color document is a color in which a document is written and a color of an instruction symbol for the document are different from each other. For example, an instruction symbol is added to a document written in black with a color pen or the like. There are things. It should be noted that any color may be used as long as the color of the document, the instruction mark, and the background color can be mutually distinguished. Reference numeral 2 is an instruction symbol cutout unit, which receives the instruction symbol binary image output from the color extraction unit 1 as an input,
The position and size of the instruction symbol are obtained and output as the instruction symbol position / size information.

Reference numeral 3 denotes an instruction symbol recognition unit, which receives the instruction symbol binary image output from the color extraction unit 1 and the instruction symbol position / size information output from the instruction symbol cutout unit 2 as inputs. For example, the instruction symbol is recognized by a technique such as pattern recognition, and is obtained and output as instruction information. Reference numeral 4 denotes a character cutout portion, which is a character 2 output from the color extraction portion 1.
The value image is input, the position and size of the character are calculated, and output as character position / size information.

Reference numeral 5 denotes a character recognition unit, which receives the character binary image output from the color extraction unit 1 and the character position / size information output from the character cutout unit 4, and, for example, pattern recognition or the like. By the method described above, a character is recognized and the character information such as a character code corresponding to the character is output. 6
Is a document configuration unit, and the instruction symbol position / size information output from the instruction symbol cutting unit 2, the instruction information cut out from the instruction symbol recognition unit 3, and the character cutting unit 4 are output. Character position / size information and the character recognition unit 5
The character information output from the above is input, and the document is reconstructed from these inputs to output the document data as, for example, text data.

FIG. 2 is an explanatory diagram of an input image. In the figure, an image of the character string "Aieo" is shown as an example of a document, and there is a symbol between the "i" and "e" that means insertion as an example of the instruction symbol. The letter "u" is shown just above. This instruction symbol is actually written in a color different from the character color, and after reading this, the insertion process is performed according to the procedure described below, and the character string of "aiueo" is reconstructed Will be output as.

First, an original is read by a color scanner or the like (not shown) to obtain a color document image. This is input to the color extraction unit 1 as multi-valued image data based on the intensities of the red component, the blue component, and the green component (hereinafter referred to as R, G, and B, respectively) in each pixel. The color extracting unit 1 compares the data of the multi-valued image, which is the input image, with the threshold values set for R, G, and B, respectively, and thereby, for each pixel, “ground color”, “character color”, "Indicator color"
It is determined to belong to any of the above.

FIG. 3 is an explanatory view of a binary character image, and FIG.
Shows an explanatory diagram of a binary image of an instruction symbol. This character 2
A value image is an image in which only the data determined to be the character color by the above-described pixel discrimination is regarded as “black” and the other than the character color is regarded as “white”, and the instruction symbol binary image is the instruction symbol. Only the data determined to be a color is regarded as “black” and the colors other than the indicator color are regarded as “white”. Although the expression “white” or “black” is used here for each pixel of the binary image, this does not indicate the actual color but can be paraphrased as “0” or “1”. It means the digitized data.

The obtained instruction symbol binary image is outputted to the instruction symbol cutting unit 2 and the instruction symbol recognition unit 3, and the character binary image is outputted to the character cutting unit 4 and the character recognition unit 5, respectively. In the character cutout unit 4, the character binary image is input, the input is subjected to processing such as character / graphic separation, and a character is cut out in an area determined to be a character to recognize the character. The position and size are obtained, the character is cut out to form a character pattern, and the character pattern is recognized by the character recognition unit 5, for example, by performing pattern recognition using a recognition dictionary in which the characteristic data of each character pattern is registered, and the corresponding The character code to be used is obtained as character information. In this way, the character position and size as well as the character information are obtained. For example, it is to obtain the position / size information and the character code of each of the character 7, character 8, character 9, character 10 and character 11 in FIG.

On the other hand, the instruction symbol cutout unit 2 receives the above-mentioned instruction symbol binary image, cuts out the instruction symbol from this input, obtains the position and size of the instruction symbol, and determines the pattern of the cut out instruction symbol. First, the instruction symbol recognition unit 3 recognizes the instruction symbol. The instruction symbol cutout and the instruction symbol recognition are performed by the same method as the above-described character cutout and character recognition, and the “instruction symbol” may be processed instead of the “character”. Therefore, similarly to the recognition dictionary in character recognition, a dedicated recognition dictionary corresponding to the instruction symbols is created and the instruction symbols are recognized.

In this way, it is possible to obtain the position and size of the instruction symbol and obtain instruction information such as a code indicating the instruction symbol. For example, it means obtaining the position / size information of the indicator 12 and the code corresponding to the indicator 12 in FIG. In the insertion symbol given as an example of the instruction symbol, it is determined that the character immediately above it is inserted after the character immediately below and to the left, and a code is given to this insertion symbol.

FIG. 5 is an explanatory diagram of position / size information.
FIG. 5A is for explaining the instruction symbol position / size information, and an insertion symbol which is an instruction symbol is shown as an example. The position and size of the instruction symbol are represented as, for example, the starting point coordinates, width, and height of a rectangle circumscribing the instruction symbol. Regarding this instruction symbol, the starting point coordinates are shown as (x1, y1), the width as w1, and the height as h1.

FIG. 5B is for explaining the character position / size information. The position and size of a character are represented as the starting point coordinates, width, and height of a rectangle circumscribing the character. For this character, the starting point coordinates are (x2, y2), the width is w2, and the height is h2. Here, as an example of the coordinate system, a description is given using a coordinate system in which the positive direction of the x-axis is the right direction in the figure and the positive direction of the y-axis is the down direction of the figure.

In the document construction unit 6, the instruction symbols are examined character by character based on the instruction symbol position / size information, instruction information, character position / size information and character information obtained as described above, and the instruction is concerned. The instruction symbol is specified by, for example, determining the code of the symbol. In the example shown in the figure, by identifying the code of the instruction symbol, it is determined that this instruction symbol is an insertion symbol, and the editing operation corresponding to this is that the character immediately above the insertion symbol is Interpreted as inserting after.

When the meaning of the instruction symbol is known, the character to be instructed by the instruction symbol is searched for next. In the present embodiment, the center coordinates (x3, y3) of the upper side of the rectangle circumscribing the indicator are obtained as x3 = x1 + w1 / 2y3 = y1 for the indicator, and the character is circumscribed with the character. The center coordinates (x4, y4) of the lower side of the rectangle are calculated as x4 = x2 + w1 / 2y4 = y2 + h2.

If any of the input characters satisfies | x3-x4 | <xs1 and | y3-y4 | <ys1, this is interpreted as the character immediately above the indicator. However, xs1 and ys1 are constants. Also, regarding the instruction symbol, the center coordinates (x5, y5) of the lower side of the rectangle circumscribing the instruction symbol are obtained as x5 = x1 + w1 / 2 y5 = y1 + h1 and regarding the character, a rectangle circumscribing the character. The vertex coordinates (x6, y6) at the upper right of is calculated as x6 = x2 + w2 y6 = y2.

If any of the input characters satisfies | x5-x6 | <xs2 and | y5-y6 | <ys2, it is interpreted as the character immediately below the left. However, xs2 and ys2 are constants. In this way, for each instruction symbol, the character immediately above and the character immediately below left are specified, and the character immediately above that is inserted after the character immediately below left. In FIG. 2, "u" is determined to be the character immediately above the insertion symbol, "i" is determined to be the character immediately below and to the left of the insertion symbol, and the character "u" is inserted after this character "i" to obtain the character string. "AIUEO" will be obtained as output document data.

In some cases, such as when inserting a character at the beginning of a character string, there is no corresponding character immediately below and to the left of the insertion symbol. In this case, the character immediately below and to the right of the insertion symbol is located immediately above the insertion symbol. You can insert a character. Also, in the example shown in the figure, horizontal writing is handled, but the character to the right of the instruction symbol is
The insertion symbol for vertical writing can be defined and registered as one of the instruction symbols by inserting it after the character directly above the instruction symbol.

In the above description, the insertion symbol is used as the instruction symbol. However, this means that the shape of the instruction symbol and the editing process can be performed by using any instruction symbol such as a symbol meaning deletion or replacement. The present invention can be similarly applied by registering the procedure of. As described above, by writing an instruction symbol of a different color from the document in the document and recognizing the instruction symbol and automatically performing the editing operation, the reading / editing process can be continuously performed without human intervention. Can be carried out by the above-mentioned method, and the processing efficiency is improved.

[0025]

As described above in detail, by writing an instruction symbol in a document and recognizing the instruction symbol and automatically performing an editing operation, the reading / reading can be performed without human intervention.
The editing process can be continuously executed, and the processing efficiency is improved.

[Brief description of drawings]

FIG. 1 is a block diagram of an embodiment.

FIG. 2 is an explanatory diagram of an input image

FIG. 3 is an explanatory diagram of a binary character image.

FIG. 4 is an explanatory view of an instruction symbol binary image.

FIG. 5 is an explanatory diagram of position / size information.

[Explanation of symbols]

 1 Color Extraction Unit 2 Indicator Symbol Extraction Unit 3 Indicator Symbol Recognition Unit 4 Character Extraction Unit 5 Character Recognition Unit 6 Document Composition Unit

Claims (2)

[Claims] 1. Input image data consisting of a character and an instruction symbol having a color different from that of the character and used for editing the character is converted into a character binary image and an instruction symbol binary image according to the color. A color extracting unit for separating, an instruction symbol cutting unit for extracting the instruction symbol position / size information from the instruction symbol binary image, and an editing process based on the instruction symbol binary image and the instruction symbol position / size information. And a character cutout unit for extracting character position / size information from the character binary image, and character information based on the character binary image and the character position / size information. Based on the character recognition unit to recognize and the above-mentioned instruction symbol position / size information, instruction information, character position / size information, and character information, edit processing corresponding to the instruction information is performed to reconstruct the document. Document structure to output Document image recognition apparatus characterized by having. 2. The document according to claim 1, further comprising a document configuration unit for performing edit processing by specifying a character to be designated by the instruction symbol based on a positional relationship between the instruction symbol and the character. Image recognition device.