JP2000020726A - Image processor and its method - Google Patents

Abstract

PROBLEM TO BE SOLVED: To make it possible to extract a character string area from a photograph or chart area or the like in a document image inputted from an image input device such as a scanner. SOLUTION: The image processor 15 provided with a character area extraction part 102 for extracting a character area from a document image inputted from the image input part 101 such as a scanner, a specific area extraction part 103 for extracting a specific area such as a photograph and a chart, a character string area extraction part 104 for extracting a character string area from a character area in the specific area, and an image discrimination part 105 for discriminating an image in the specific area by using the extracted character string information.

Description

DETAILED DESCRIPTION OF THE INVENTION

[0001]

BACKGROUND OF THE INVENTION 1. Field of the Invention The present invention relates to an image processing apparatus and method for converting a density value or the like when filing or copying an image such as a document.

[0002]

2. Description of the Related Art Techniques for capturing a document image as image data and outputting the image as a hard copy or saving the image file as an image file include a technique used for image capture in a copying machine and a personal computer, a filing system and the like. There are technologies used for databases.

[0003] In the case of outputting as a hard copy, image processing is sometimes performed to make it easy for a user to view as a document image. For example, when a character is included in the document image, the edge portion of the character is emphasized by performing the high-pass filter processing, the output becomes clear, and the character becomes easy to read.

In the case of a photograph, a low-pass filter is applied to reproduce a smooth gradation, so that a rough feeling is eliminated and a beautiful output is obtained. If the same photo is a halftone dot photo,
It is also possible to switch the processing depending on whether the photograph is a silver halide photograph. As described above, changing the processing method according to the type of the document image is very effective both when outputting as a hard copy and when saving as an image file.

[0005] A document image is usually composed of areas such as characters, figures, photographs, etc., singly or overlapping. Therefore, when performing image processing, it is necessary to detect what area and how are configured in the document image.

Conventionally, as disclosed in Japanese Patent Application Laid-Open No. 9-92450, layout analysis has been used to identify elements constituting a document image such as characters, halftone dots, and photographs.
There is a method for extracting the structure of a document image. in this way,
After performing the binarization process, the connectivity of pixels is checked, connected pixels are extracted as regions, and identification is performed using feature amounts such as positions and sizes. In the case of this method, since binarization is performed as preprocessing, it has been difficult to accurately extract a region including an intermediate density value such as a photograph.

Further, as described in Japanese Patent Application No. 10-053317 as an extension of this method, a multi-valued image is converted into a plurality of binary images, and a layout is applied to each of the binary images. In some cases, an area including an intermediate density value such as a photograph is accurately extracted by performing analysis.
This method has an advantage that the region can be correctly extracted even when the regions such as characters on the halftone dot overlap.

However, an actual document original often has a complicated layout structure. For example, characters are written on the area where the halftone picture exists. In this case, in the method described in Japanese Patent Application Laid-Open No. 9-92450, since binarization is performed, it is difficult to separate the halftone photograph and the character by the density value, and the area cannot be correctly extracted.

In the method described in Japanese Patent Application No. 10-053317, it is identifiable that a photographic area exists on a halftone dot area.
It is difficult to separate a character from a halftone dot photograph because a feature amount such as a size of a connected component in a photograph region cannot be distinguished from a feature amount for extracting a character region.

[0010]

As described above, in a conventional image processing apparatus, a multi-valued image is converted into a plurality of binary images, and each binary image is subjected to layout analysis. In the method of extracting the structure of a document image, it is difficult to separate a correct character when extracting a character region existing on a halftone dot photograph. Accordingly, it is an object of the present invention to provide an image processing apparatus for accurately extracting characters in a region such as a photograph from document image data.

[0011]

In order to achieve the above object, the present invention provides a character area extracting means for extracting a character area having a character characteristic from the density values of each pixel and neighboring pixels of an image read as image data. A specific area extracting means for extracting a specific area having characteristics other than characters, and a character string using information on the character area extracted by the character area extracting means from the specific area extracted by the specific area extracting means. A character string region extracting means for extracting an area, and an image determining means for determining an image of the specific area using information on the character string region extracted by the character string region extracting means.

Further, a character region extracting means for extracting a character region having character characteristics based on density values of pixels and neighboring pixels of an image read as image data, and a specification for extracting a specific region having characteristics other than characters A region extracting unit, a character string region extracting unit that extracts a character string region from information of the character region extracted by the character region extracting unit from the specific region extracted by the specific region extracting unit, A document structure storing means for storing the position of the character string area as a document structure, and comparing the character string area extracted by the character string area extracting means with the document structure stored in the document structure storing means. Character string collating means to be determined, and image presenting means for presenting an image read using information on the character string area determined by the character string collating means. And wherein the Rukoto.

That is, when a document is copied or converted into an image and stored, the change in pixel value and the characteristic amount of the input document image are measured so that the document element included in the image is defined as an area. Extracting, identifying document elements for each area, and extracting character strings from information such as the position and size of character areas existing in specific areas such as photographs, etc. It is possible to correct the pixel attribute of a specific area such as a photograph using character string information. Further, by comparing the document structure pattern with the character string information, an image search can be performed.

[0014]

DESCRIPTION OF THE PREFERRED EMBODIMENTS In the present invention, a document image is captured by an image input unit such as a scanner, and the character area extracting means and the area extracting means physically or logically determine the state of the data of each pixel and neighboring pixels. After extracting what is connected as a single area, the position, size, shape, structure, density distribution, and other features of each area on the image are measured, and the measurement results are determined in advance. Based on the rule, it is identified as an area having a specific attribute such as a character or a photograph as a document constituent element.

Then, for a region having a character attribute existing in a region having a specific attribute such as a photograph, a characteristic amount such as an attribute, a size, and a shape of another region existing around the region is measured. As a result, if it is determined that the area having the corresponding character attribute is an element constituting the character string, a new pixel attribute can be given as the character string area, and the pixel attribute of the area having the specific attribute such as a photograph is corrected. It is possible to do.

Hereinafter, a more specific embodiment of the present invention will be described with reference to the drawings. FIG. 1 is a block diagram showing an embodiment of the present invention.

An image input unit 101 is a device for inputting image data. An image input device such as an image scanner, which is a device for reading a document and converting it into image data, captures an image drawn from a document such as a document. Device. Note that the image input unit in the present invention may be configured by a reading device such as an image scanner, or may be configured by a device that captures an image already stored in a filing device or the like as image data.

The character area extraction unit 102 is provided with the image input unit 10
In step 1, pixels having characters from the input image data are used as regions. First, the image input unit 101
The input image data is separated into a plurality of binary image data according to the state such as the density difference and the saturation of peripheral pixels, and characters or graphics are physically or logically connected from the image. The area is extracted as a document element by measuring features such as the position, size, shape, structure, and density distribution of the area.

As a specific method of separating the image data into a plurality of binary image data, a method disclosed in Japanese Patent Application No. 10-053317 may be used.

In this method, a multi-valued image is converted into a plurality of binary images, and a layout analysis is performed on each of the binary images. Is to be extracted. In this case, seven binary separated image data of a character image, a halftone image, a background image, a halftone image, a color image, a gray image, and a black image are generated.

Further, as a specific method of extracting and identifying document elements, a method disclosed in Japanese Patent Application Laid-Open No. 9-92450 may be realized.

In this method, after performing a binarization process, the connectivity of pixels is checked, connected pixels are extracted as regions, and identification is performed using feature amounts such as the position and size. Things.

The specific area extracting unit 103 includes the image input unit 10
In step 1, a pixel having a specific attribute is extracted as an area from the input image data. Examples of the attribute in this case are a photograph, a chart, and a dot area. Similarly to the character area extraction unit 102, the input image data is separated into a plurality of binary image data according to the state such as the density difference and the saturation between each pixel and the surrounding pixels, and characters or characters are extracted from each image. Extract regions where figures etc. are physically or logically connected and extract the position, size, shape,
It measures features such as structure and density distribution, and identifies each type and importance as document elements.

The types of document elements include photographs, figures, tables, and halftone dots. The specific area extraction unit 103 integrates not only single image data but also feature amounts of a plurality of pieces of image data according to rules, and determines an area attribute of input image data. For example, when an area is extracted at the same position from both the character image and the halftone image, the type of the area and the size of the area are determined. As a specific example, when a photographic area exists on the character image and a halftone pixel exists in the halftone image at the same position, it is determined to be a silver halide photographic area.

As a specific method for determining an area attribute from a plurality of image data, it may be realized by the method disclosed in Japanese Patent Application No. 10-053317. The character string region extracting unit 104 will be described in detail later, but the specific region extracting unit 1
A character string existing in an area such as a photograph, a figure, a table, or a halftone dot extracted in 03 is extracted. At this time, a photograph,
Measure the features, such as attributes, size, shape, etc. of other surrounding areas for those existing in areas such as figures, tables, halftone dots, etc. If it is determined that the element forms a character string, it is extracted as a character string area.

The details of the image discriminating unit 105 will be described later.
Using the information on the character string area extracted by the character string area extraction unit 104, the area information of the attributes such as a photograph, a figure, a table, and a halftone dot output from the specific area extraction unit 103 is corrected.

The image output unit 106 outputs according to the area information determined by the image determination unit 105.
As an output form, the area information of the attribute may be converted into an image. As a specific method of converting the attribute area information into an image, for example, the method may be realized by a method disclosed in Japanese Patent Application No. 10-053317.

FIG. 2 is a flowchart showing an example of a document processing procedure in the configuration of the apparatus of the present invention shown in FIG.
Next, the flow of processing of the image processing apparatus according to the present invention will be described with reference to this flowchart and the configuration diagram of FIG.

First, an image of a document is captured by the image input unit 101 (step ST201). That is, the image input unit 101 converts an image read from a document using an image input device such as a scanner or image file data input from a filing system or the like into image data.

In the character area extracting unit 102, one line or several lines are read from the image input device.
It is separated into a plurality of binary image data according to the state of each pixel such as the density difference and the saturation of the peripheral pixels, and is divided and extracted for each area where characters and figures are physically or logically connected. Then, the feature amounts such as the position, size, shape, structure, and density distribution of the area are measured, and the type and importance of the area are identified. As a result, only the character area is extracted (step ST202). The processing is repeated until the processing is completed for all the pixels of the input image (step ST203).

Similarly, in the specific area extracting unit 103,
One line or several lines are read from the image input device, and separated into a plurality of binary image data according to the state of each pixel such as the density difference and saturation of peripheral pixels. Characters and figures are physically or logically read. It extracts and divides each of the regions that are connected to each other, measures the position, size, shape, structure, density distribution, etc. of the region, and identifies the type and importance of the region. Then, only the photograph, the figure, the table, and the halftone dot area are extracted (step ST204). The processing is repeated until the processing is completed for all the pixels of the input image (step ST205).

After all the pixel attributes have been determined, the character string region extraction unit 104 sets the attribute of the region having the character attribute existing in the region of the specific region extraction unit 103 to the attribute of another region existing therearound. , Size, shape, and the like, and as a result, when it is determined that the area having the corresponding character attribute is an element constituting the character string, a new pixel attribute is given as a character string area (step ST206). The processing is repeated until the processing ends for the character attributes in all the specific areas (step ST207). Note that the character string region extraction unit 10
4 will be described later in detail with reference to FIG.

After the character string region extraction unit 104 extracts the character string regions from all the specific regions, the image is determined using the character string information for each specific region (step ST208). As an example of the discrimination, an area of only a photograph,
There are an area for photographs and characters, an area for figures and characters, and the like. The information of the pixel attribute is changed based on the result of this determination. And
The processing is repeated until the processing is completed for all the specific areas (step ST209).

Finally, the image output unit 106 outputs an output image obtained by converting each pixel attribute as image data (step ST210). The above is a rough processing operation of the apparatus of the present invention. Next, details of processing of each element will be described.

FIG. 3 is a flowchart showing details of an example of a character string extraction process in the character string region extraction unit 104 of the image processing apparatus of the present invention, and is a flowchart of the process performed in step ST208 shown in FIG. is there.

In the character string extraction process, a region having the character attribute is examined by examining the relationship between the region having the character attribute and other pixels having the character attribute around the region. This is a process for examining whether a character string area is finally extracted. Hereinafter, description will be given according to the flowchart.

Character region extraction unit 102 and specific region extraction unit 1
The pixel attribute data 151, which is an output from the image data 03, determines the attribute of each pixel of the input image data by using the type and importance of the area of the identified plurality of image data. The attributes of the specific area from which the character string is to be extracted include
There are halftone dot attributes extracted from a halftone dot image, such as a photograph attribute, a drawing attribute, and a table attribute extracted from a character image and a halftone image. In the character string extraction processing, the character string area is extracted by using the character areas existing in these attribute areas as clues.

First, data is read from the pixel attribute data 151, and it is checked whether the data is a photograph attribute, a drawing attribute, a table attribute, or a halftone dot attribute which is a specific area (step ST3).
01). When the specific area is a photograph attribute, a drawing attribute, a table attribute, and a halftone dot attribute, it is checked whether an area having a character attribute exists in the area. (Step ST302). When there is a region having a character attribute, the region is searched to check the state of the surrounding pixel attribute. For this purpose, a search area is set (step ST303).

FIG. 4 shows an example of a method for setting a search area from a character attribute area. The coordinate values of the hatched area 501 determined to have the character attribute in step ST302 are (x0, y0) and (x1, y1). The coordinates (X0, X0) of the dotted line area 502 surrounding the character attribute area 501
Y0) and (X1, Y1) are obtained from the following equations. X0 = x0− (x1−x0) × m Y0 = y0− (y1−y0) × m X1 = x1 + (x1−x0) × m Y1 = y1 + (y1−y0) × m

Here, m is a coefficient for setting the search area, and may be a constant or may be variable depending on the size of the character attribute area 501. From the area 502 and the character attribute area 501, a cross-shaped search area 503 surrounded by a thick line centering on the character attribute area 501 is set.
The reason why the search area is set to the cross shape is to take into account the characteristics of the character string.

That is, if the character attribute area 501 is a character constituting a character string, it is considered that an area having the same character attribute exists on the left, right, up, or down. In addition, by setting the search area in a cross shape, it is possible to perform a search regardless of whether the direction of the character string is vertical or horizontal.

Next, a search is performed to determine whether there is an area having another character attribute existing in the set search area 503 or overlapping the search area 503 (step ST30).
4). In FIG. 4, a region 504, a region 505, and a region 5
06 and the area 507 are areas having character attributes. Regions existing within the search region 503 or satisfying the condition of overlapping with the search region 503, that is, the regions 504, 505, and 506 are the corresponding regions. The overlapping area is also made effective because it does not always fit in the cross-shaped search area 503 due to the inclination of the document.
Therefore, even if the input document image has a slight inclination, it can be detected without any problem.

In step ST304, if there is a region having a character attribute, a vote is made in a voting space in which the rectangular size of the region having the character attribute is prepared in advance (step ST305). This voting space is used for extracting a character string area. In the method of directly extracting a character string by examining the positional relationship between a character attribute region serving as a clue and a region having a detected character attribute existing in a search range, the position is determined as the number of character attribute regions increases. The processing cost for evaluating the relationship increases and is not suitable for high-speed processing.

Therefore, a character attribute area which is presumed to be a component of a character string is voted in the voting space, and later extracted as a character string area for each specific area unit. The voting space prepared here is assumed to be lower than the resolution of a plurality of binary images.

Further, two types, one for the vertical direction and the other for the horizontal direction, are prepared according to the direction of the character string, and the voting space for the vertical direction has a lower resolution in the vertical direction than in the horizontal direction. The voting space has lower horizontal resolution than vertical resolution. This makes it easier to connect character attribute regions, which are constituent elements of a separated character string by lowering the resolution.

Also, by separating the vertical and horizontal directions,
It is possible to deal with cases where the character string direction is unknown or mixed. For these reasons, extraction of a character string area from a voting space described later (step ST306)
, Easy character string extraction is possible.

The voting is performed by adding a feature such as the rectangular size of the area 501 to a position where the coordinate position of the area 501 having the character attribute is converted to the coordinate position of the voting space. As an example of the rectangle size, any of the width, height, area value, and the like of the rectangle can be used. At this time, depending on the distribution state of the area having the character attribute detected in step ST304, whether to vote in the vertical voting space, in the horizontal voting space, or in both voting spaces To determine.

That is, when there is a region having a character attribute detected in the vertical direction, a vote is cast in the vertical voting space,
If there is a region having a character attribute detected in the horizontal direction, the ballot is cast in the horizontal voting space. Taking FIG. 4 as an example,
Character attribute areas detected for the area 501 having the character attribute are an area 504, an area 505, and an area 506,
It exists both in the vertical and horizontal directions. Therefore, in this case, votes are given to both voting spaces.

Further, in the cross-shaped search area 503,
The region having the detected character attribute is also a target of voting. Those detected in the horizontal direction are voted in the voting space in the horizontal direction, and those detected in the vertical direction are voted in the voting space in the vertical direction.

Referring to FIG. 4 as an example, the region 504 and the region 50
6 is cast in the horizontal voting space, and the area 505 is cast in the vertical voting space. These voting is performed by adding a feature such as a rectangular size of the area to a place where the coordinate position is converted into a coordinate position in the voting space, similarly to the voting in the area 501.

After all rectangular-size voting of the region having the character attribute to the voting space is completed, or
If there is no area having the character attribute in step 304, the process returns to step ST302, and the process returns from step ST303 to step ST30 for the area having another character attribute.
Step 5 is performed.

When the process is completed for all the regions having the character attribute in the specific region, a character string region is extracted from the voting space (step ST306). Step ST304
Can be regarded as a multi-valued image.

Therefore, the voting space is converted into a binary image by threshold processing. This threshold processing has the following meaning. In step ST304,
Votes are also made for areas that have character attributes despite being noise components. However, at locations where the probability of the presence of a character string is high, that is, at locations where there are many regions with character attributes, the voted value is high. Conversely, locations voted by noise components are due to other regions. The value is low because no voting is done.

Therefore, at the same time as the conversion into the binary image by the threshold processing, the noise component unnecessary as the character string area is deleted. The setting of the threshold may be a constant,
It may be calculated from the voting value in the voting space.

The method of extracting a character string area from the binarized voting space that has been subjected to the threshold processing is performed by the character string area extracting unit 1
The same method as the extraction of the connected component in 04 may be used.

After the end of the character string area extraction, the processing from step ST302 to step ST306 is repeated for other specific areas. As a result of performing the process on all the specific areas, the extracted information on the character string area becomes corrected pixel attribute data 152 as a result of adding the attribute of the character string area to the pixel attribute data 151.

FIG. 5 shows an example of image data input to the image processing apparatus of the present invention. Here, image data 6
01 indicates a halftone dot photograph 602 and a silver halide photograph 603, which are photographs of a camera including a character string, and a character string 604 on the background. The actual processing in the character string region extraction unit 104 will be described using this example.

A character area extracting unit 102 extracts a character area 602 indicated by a dotted line in the image data 601a in FIG. 6 from the image data 601 input from the image input unit 101.
a1, 602a2, 603a1, 603a2, 604a
Is extracted as a rectangular area.

For convenience, a plurality of character areas are surrounded by dotted lines, but actual character areas are individually extracted as small rectangles. Here, character areas 602a1 and 603a1
Are the dot photograph 602 and the silver halide photograph 603 in FIG.
The size and shape are similar to those of the character attribute, and are not correct character areas.

Similarly, from the image data 601 input from the image input unit 101, the specific area extracting unit 103 converts the halftone dot area 602b indicated by the dotted line in the image data 601b in FIG. Is extracted as

In the character string region extracting unit 104, the halftone dot region 602b extracted in FIG. 7 and the character regions 602a1, 602a2, 603a1, and 6 existing in the silver halide photographing unit 603b.
Character string extraction is performed on 03a2. Character area 602
a1 and 603a1 are not extracted as a character string area because there is no character attribute in the vicinity in the search for the character attribute or the value in the voting space is small even if a rectangle-sized voting is performed. The character string area 602c2 indicated by the image data 601c is extracted from the vertical voting space, and the character string area 603c2 is extracted from the horizontal voting space as a character string. The above is the description of the actual processing in the character string region extraction unit 104.

Next, a description will be given of the character string area extracting means according to claim 4 of the present invention. A flowchart showing details of a character string extraction process as an example in the present invention is the same as that in FIG.

It is checked from the pixel attribute data 151 whether an area having a specific attribute exists, and it is checked whether an area having a character string attribute exists in the specific quantity machine. Then, a search area is set, and the search area is set. The characters within the character space are searched, the rectangular size of the character attribute area detected as a result is voted in each of the voting space in the vertical direction and the horizontal direction, and the voting space is converted into a binary image by threshold processing. Extract the column area.

Further, a paragraph is extracted by measuring a positional relationship between the extracted character string area and another character string area. By determining whether the character string is a character string from the position information of the character string area extracted from the paragraph information, it is possible to more accurately extract the character part.

FIG. 9 shows an example of image data input to the image processing of the present invention. Here, image data 701
Shows a halftone dot picture 702 which is a picture of a camera including a character string of a plurality of lines. Using this example, a method of extracting paragraph information from the extracted character string area information and a method of limiting the character string area from the information will be described.

From the image data 701 input from the image input unit 101, the character area extracting unit 102 sets the character area 70 indicated by a dotted line in the image data 701a in FIG.
2a1 and 702a2 are extracted as rectangular areas. Here, the character area 702a1 is the halftone picture 702 in FIG.
Are separated from the camera part, and their size and shape indicate an area similar to the character attribute, and are not actually correct character areas.

Similarly, from the image data 701 input from the image input unit 101, the dot area 702b indicated by the dotted line in the image data 701b in FIG. The character string extraction is performed by the region extracting unit 104 on the character regions 702a1 and 702a2 existing in the dot region 702b. As a result, a character string area 702 shown in FIG.
c1 and 702c2 are extracted as character strings from the horizontal voting space. For convenience, 702c2 surrounds a plurality of character string regions by dotted lines, but actual character string regions are extracted as individual regions.

Here, by measuring the positional relationship between the extracted character string area and other character string areas, the character string area 702c2 shown by the image data 701c in FIG. Extracted as a paragraph. If only a region having a paragraph shape is extracted as a character string region as a setting of the system, a false character string region 702c1 separated from the camera portion of the halftone dot photograph 702 is used as a character string. Are not extracted, and finally the character string area 702d2 shown in FIG.
Will remain as a character string.

As described above, the document structure of the input document image is extracted from the extracted character string area information, and the character string area extracted using the document structure is extracted. By checking again, it is possible to extract a character string with higher accuracy. The above is the description of the embodiment of the present invention shown in FIG.

Next, an apparatus for extracting a character string region with higher accuracy by collating the extracted character string region information with document structure information and an image search device as an application example thereof will be described.

FIG. 14 is a block diagram showing an example of the second embodiment of the present invention. FIG. 14 shows an image input unit 101, a character region extraction unit 102, a specific region extraction unit 103, a character string region extraction unit 104, a document structure storage unit 107, and a character string collation unit 1.
08, an image determination unit 105, and an image output unit 106,
In this embodiment, a document structure storage unit 107 and a character string collation unit 108 are added to the block diagram of FIG.

Image input unit 101, character area extraction unit 10
2. Specific area extraction unit 103, character string area extraction unit 104,
Since the image discriminating unit 105 and the image output unit 106 have already been described, only the document structure storage unit 107 and the character string matching unit 108 will be described here.

FIG. 15 is a flowchart showing an example of a document processing procedure in the configuration of the apparatus of the present invention shown in FIG. The flowchart of the configuration in FIG. 1 shown in FIG. 2 includes a character string collation process (step ST211 and step S211).
T212).

After the character string extraction is completed (step ST
207), and a character string collation process (step ST211) is performed. Here, the character string area is collated using the extracted character string area information and the document structure pattern previously given to the document structure storage unit 107 as knowledge data. Determine as area. The document structure pattern given as the knowledge data may be, for example, data indicating a coordinate position of a rectangular area, or convert script information indicating a document structure such as "character string under a photograph area" into pattern information. Data or a simple figure showing a region to be extracted may be input by a scanner. When the data is input by the scanner, the document structure is extracted from the figure and converted into data.

First, the position, size, and the like are obtained from the extracted character string area information. Next, the position and size of a character string to be extracted are acquired from the document structure pattern of the given knowledge data. Next, each character string information is collated.
At this time, if the size of a specific area such as a photograph is different, the size is adjusted by normalization.

Then, the degree of overlap is represented by the degree of similarity, and if the degree of similarity is equal to or greater than the threshold value, it is considered that collation has been performed. If the character string area indicated in the knowledge data exists as a result of the collation, the area is determined as a character string, and if there is any other character string area, information for discrimination is used as character string area information. To be added. It repeats until the collation processing is completed for all the character string data of the document structure information of the knowledge data (step ST212). After finishing,
The image discriminating unit 105 performs a discriminating process. At this time, the result of the character string collating unit 108 is compared with the character string region extracting unit 1.
By regarding this as the output of 04, the same processing as in the example shown in FIG. 1 can be performed.

FIG. 16 shows an example of image data input to the image processing of the present invention. Here, the image data 80
1, a halftone dot photograph 802 which is a photograph of a camera including a character string of only one line is shown. The operation of the character string matching unit 108 will be described using this example.

The image data 801 input from the image input unit 101 is extracted by the character area extracting unit 102 into the character area 80 indicated by a dotted line in the image data 801a in FIG.
2a1 and 802a2 are extracted as rectangular areas. Here, the character area 802a1 is the halftone picture 80 in FIG.
The camera is separated from the camera part No. 2 and its size and shape indicate an area similar to the character attribute, and is not actually a correct character area.

Similarly, from the image data 801 input from the image input unit 101, the dot area 802b indicated by the dotted line in the image data 801b in FIG. The character string extraction is performed by the region extracting unit 104 on the character regions 802a1 and 802a2 existing in the halftone dot region 802b. As a result, the character string area 802 shown in FIG.
c1 and 802c2 are extracted as character strings from the horizontal voting space.

Here, the document structure is extracted by measuring the positional relationship between the extracted character string area and other character string areas. In the configuration example of the present invention shown in FIG. 1, a paragraph is extracted and a character string area is extracted based on the result. On the other hand, the character string of the image data 801 is composed of only one line. The paragraph cannot be extracted, and only the character string area 802c2 cannot be extracted.

FIG. 20 shows an example of a document structure pattern given to the document structure storage unit 107 according to the present invention. Here, the image data 901 specifies position information for the region 902 and the character string region for the region 903, respectively. The character string area is determined by comparing the document structure pattern of this data with the result of FIG. As a result, the character string area 802c2 in FIG.
Corresponds to the character string area 903, the character string 802d2 shown in FIG. 21 can be left as a character string area.

As described above, by collating the extracted character string area information with the character string area in the document structure pattern given in advance, it is possible to extract a character string according to the purpose with high accuracy. Becomes The preceding is an explanation of the embodiment of the present invention shown in FIG.

Further, in another form, the present invention can be used as an apparatus for performing image retrieval.

FIG. 22 is a block diagram showing an example of the third embodiment of the present invention. FIG. 22 shows an image input unit 101, a character region extraction unit 102, a specific region extraction unit 103, a character string region extraction unit 104, a document structure storage unit 107, and a character string collation unit 1.
08, and an image presentation unit 109. The image discriminating unit 105 and the image output unit 106 are removed from the block diagram of FIG. 14, and an image presenting unit 109 is added.
Since the image input unit 101, the character area extraction unit 102, the specific area extraction unit 103, the character string area extraction unit 104, and the document structure storage unit 107 have already been described, here, the character string collation unit 108 and the image presentation unit 109 Will be described.

It is assumed that the image data is a document image stored in a filing device or the like. As described above, the character string matching unit 108 performs character string matching based on the extracted character string area information and the document structure pattern given in advance to the document structure storage unit 107, and sets the area as a character string area. It will be decided.

Therefore, when the user specifies the document structure pattern of the image image to be searched for in the document structure storage unit 107, the collation result in the character string collation unit 108 becomes a search result. If the result of the collation by the character string collating unit 108 indicates that the information matches, the image presenting unit 10 regards the image data as desired by the user.
9 is presented with the image.

The image presenting unit 109 may display an image like a display, or output it to paper like a printer. By performing the above processing on a plurality of image data, a retrieval operation of a document image in the filing apparatus is performed. As described above, it is possible to search for image data by comparing the information on the extracted character string area with the character string area extracted from the document structure pattern of the image image that the user wants to search. The preceding is an explanation of the embodiment of the present invention shown in FIG.

As described above, in the present apparatus, the image data is fetched by the image input unit 101,
2 and the specific area extraction unit 103 extracts the physical or logical connection of the characteristics of each pixel and neighboring pixels as one area, and then positions, sizes, shapes, and the like of the individual areas on the image. A characteristic amount such as a structure or a density distribution is measured, and the measurement result is identified as a region having a specific attribute such as a character or a photograph as a document component based on a predetermined rule.

Then, in the character string region extraction unit 104, for the region having the character attribute existing in the region having the specific attribute such as a photograph, the attribute, size, shape, etc. of the other region existing around the region. If the area having the corresponding character attribute is determined to be an element constituting a character string, extraction is performed as a character string area, and a specific attribute area is determined by the image determination unit 105. The pixel attribute of the area can be corrected.

Further, the document structure pattern previously given as knowledge data to the document structure storage unit 107 and the extracted character string region information are collated by the character string collation unit 108, and the character string region is determined, so that the character string region is determined. This allows for easy extraction of character strings. Further, the character string matching unit 108
The image search is made possible by using the result of the collation in.

Therefore, in the document image data input by the present apparatus, a character area is distinguished from a different document element such as a photograph, a figure, a table, a halftone dot and the like, and the area is extracted. By extracting a character string from, it becomes possible to accurately separate a character similar to a character from a character, and by using the information of this character string to correct the pixel attribute of a region such as a photograph, Compatible with complex area determination.

By using this output result for image processing such as filtering and compression, a beautiful output can be obtained even when outputting as a hard copy, and characters and photographs can be separated even when saving as an image file. It is also possible to do.

Further, by collating a given document structure pattern with character string information, it is possible to search for desired data from the filed image data.

[0094]

As described above, according to the present invention, when a document is to be hard-copied or converted into image data and stored, the text area, the photograph, By extracting and identifying different document elements such as tables, halftone dots, etc. as areas, and extracting character strings from the positions of character areas existing in areas such as photographs, figures, tables, halftone dots, etc. Similar parts and characters can be accurately separated. By utilizing the information of this character string, accurate determination can be made for a complicated area such as a photograph containing characters.

As a result, for example, when a hard copy is taken, it is possible to apply a filtering process so that the characters present in the photograph are easy to see, and to separate the character and the photograph when saving the image file, and to reuse the photograph. Easy image data can be obtained.

Further, by comparing a given document structure pattern with character string information, data desired by the user can be searched from the filed image data.

[Brief description of the drawings]

FIG. 1 is a configuration diagram of an image processing apparatus according to the present invention;
FIG. 2 is a block diagram showing an example of the embodiment.

FIG. 2 is a flowchart showing an example of image processing of the apparatus of the present invention having the configuration of FIG. 1;

FIG. 3 is a flowchart showing a character string extraction process of the image processing apparatus according to the present invention.

FIG. 4 is an explanatory diagram of a search area setting and search method of a character attribute in a character string extraction process of the image processing apparatus according to the present invention.

FIG. 5 is a view showing an example of a document image handled in the image processing apparatus according to the present invention.

FIG. 6 is an output example of a character area in a character area extraction process of the image processing apparatus according to the present invention.

FIG. 7 is an output example of a specific area in a specific area extraction process of the image processing apparatus according to the present invention.

FIG. 8 is an output example of a character string area in the character string area extraction processing of the image processing apparatus according to the present invention.

FIG. 9 is a view showing an example of a document image handled by the image processing apparatus according to the present invention.

FIG. 10 is an output example of a character area in a character area extraction process of the image processing apparatus according to the present invention.

FIG. 11 is an output example of a specific area in a specific area extraction process of the image processing apparatus according to the present invention.

FIG. 12 is an output example of a character string area in the character string area extraction processing of the image processing apparatus according to the present invention.

FIG. 13 is an output example of a character string area in the character string area extraction processing of the image processing apparatus according to the present invention.

FIG. 14 is a block diagram of an image processing apparatus according to the present invention, and is a block diagram showing an example of an embodiment.

FIG. 15 is a flowchart showing an example of image processing of the apparatus of the present invention having the configuration of FIG. 14;

FIG. 16 is a view showing an example of a document image handled in the image processing apparatus according to the present invention.

FIG. 17 is an output example of a character area in a character area extraction process of the image processing apparatus according to the present invention.

FIG. 18 is an output example of a specific area in a specific area extraction process of the image processing apparatus according to the present invention.

FIG. 19 is an output example of a character string area in the character string area extraction processing of the image processing apparatus according to the present invention.

FIG. 20 is a diagram showing an example of a document structure given to the image processing apparatus according to the present invention.

FIG. 21 is an output example of a character string area in the character string area extraction processing of the image processing apparatus according to the present invention.

FIG. 22 is a configuration diagram of an image processing apparatus according to the present invention, and is a block diagram illustrating an example of an embodiment.

[Explanation of symbols]

 101 image input unit 102 character region extraction unit 103 specific region extraction unit 104 character string region extraction unit 105 image discrimination unit 106 image output unit 107 document structure storage unit 108 character string collation unit 109 image presentation Department

 ──────────────────────────────────────────────────の Continued on the front page F term (reference) 5B057 CA16 CC03 CH11 DA08 DB02 DC30 DC36 5L096 AA06 BA07 BA08 BA20 CA18 DA01 DA05 EA35 EA43 FA02 GA07

Claims (12)

[Claims] 1. A character region extracting means for extracting a character region having a character characteristic based on the density values of each pixel and neighboring pixels of an image read as image data, and a specification for extracting a specific region having a characteristic other than the character. Area extracting means; character string area extracting means for extracting a character string area from the specific area extracted by the specific area extracting means using information on the character area extracted by the character area extracting means; An image processing apparatus comprising: an image determining unit that determines an image of the specific region using information on a character string region extracted by the region extracting unit. 2. The image processing apparatus according to claim 1, wherein the specific area extracted by the specific area extracting means is at least one of a photograph, a figure, a table, and a halftone dot. 3. The image processing apparatus according to claim 1, wherein the character string area extracting means extracts a vertical character string area and a horizontal character string area from the image data individually. 4. The image processing apparatus according to claim 1, wherein said character string area extracting means extracts a paragraph area by measuring a positional relationship between a plurality of character string areas. 5. A character region extracting means for extracting a character region having a character characteristic based on the density values of each pixel and neighboring pixels of an image read as image data, and a specification for extracting a specific region having a characteristic other than the character. Region extracting means; and character string region extracting means for extracting a character string region from information of the character region extracted by the character region extracting means from within the specific region extracted by the specific region extracting means. A document structure storing means for storing the position of the character string area as a document structure, and comparing the character string area extracted by the character string area extracting means with the document structure stored in the document structure storing means,
An image comprising: character string matching means for determining a character string area; and image determining means for determining an image of the specific area using information on the character string area determined by the character string matching means. Processing equipment. 6. A character region extracting means for extracting a character region having a character characteristic based on the density values of each pixel and neighboring pixels of an image read as image data, and a specification for extracting a specific region having a characteristic other than the character. Region extracting means; and character string region extracting means for extracting a character string region from information of the character region extracted by the character region extracting means from within the specific region extracted by the specific region extracting means. A document structure storing means for storing the position of the character string area as a document structure, and comparing the character string area extracted by the character string area extracting means with the document structure stored in the document structure storing means,
Image processing comprising: character string matching means for determining a character string area; and image presenting means for presenting an image read using information on the character string area determined by the character string matching means. apparatus. 7. A character region having character characteristics and a specific region having non-character characteristics are extracted based on density values of each pixel and neighboring pixels of an image read as image data, and from the extracted specific regions, An image processing method, wherein a character string region is extracted using the extracted character region information, and an image of the specific region is determined using the extracted character string region information. 8. A character region having character characteristics and a specific region having non-character characteristics are extracted based on the density values of each pixel and neighboring pixels of an image read as image data, and from the extracted specific regions, Extracting a character string area using the extracted information on the character area; collating the extracted character string area with a document structure indicating the position of the character string area to be extracted to determine the character string area; An image processing method, wherein an image of the specific area is determined using information of the determined character string area. 9. A character region having character characteristics and a specific region having non-character characteristics are extracted based on the density values of each pixel and neighboring pixels of an image read as image data, and from the extracted specific regions, Extracting a character string area using the extracted information on the character area; collating the extracted character string area with a document structure indicating the position of the character string area to be extracted to determine the character string area; An image processing method characterized by presenting an image read using information of a determined character string region. 10. A machine-readable recording medium on which a program for performing image processing is recorded, wherein a character area and a character other than a character having character characteristics are determined by the density values of each pixel and neighboring pixels of an image read as image data. Extracting a specific region having the following characteristics: extracting a character string region using information of the character region extracted from within the extracted specific region; and specifying the specific region using information of the extracted character string region. A recording medium on which a program for determining an image of an area is recorded. 11. A machine-readable recording medium on which a program for performing image processing is recorded, wherein a character area and a character other than a character having character characteristics are determined by density values of each pixel and neighboring pixels of an image read as image data. Extracting a specific region having the following characteristics: extracting a character string region using information of the character region extracted from the extracted specific region; extracting the extracted character string region into a character string to be extracted A recording medium on which a program for determining a character string area by collating with a document structure indicating a position of the area and determining an image of the specific area using information of the determined character string area is recorded. 12. A machine-readable recording medium on which a program for performing image processing is recorded, wherein a character area and a character other than a character having a character characteristic are represented by density values of pixels and neighboring pixels of an image read as image data. Extracting a specific region having the following characteristics: extracting a character string region using information of the character region extracted from the extracted specific region; extracting the extracted character string region into a character string to be extracted A recording medium storing a program for determining a character string region by comparing it with a document structure indicating the position of the region and presenting an image read using information of the determined character string region.