JP3020293B2 - Attribute determination method - Google Patents

Description

DETAILED DESCRIPTION OF THE INVENTION

[0001]

BACKGROUND OF THE INVENTION 1. Field of the Invention The present invention relates to a document image for identifying attributes such as a photograph area and a chart area in a document image in a communication device such as a facsimile, a document image database input device, an optical character reading device (OCR), or the like. The attribute determination method.

[0002]

2. Description of the Related Art Conventionally, this type of attribute discrimination method includes:
For example, there is one described in JP-A-62-71379.

In this document image attribute discriminating method, document image data is input, black pixels are counted (counted) in a main scanning direction (for example, a horizontal direction), and lines exceeding a threshold value are detected. A state in which a predetermined number of white lines equal to or less than the threshold value continues is determined, and the first area is cut out. In the first region cutout, the sub-scanning direction (for example, the vertical direction)
, A row in which the count exceeds the threshold is detected, and a state in which a predetermined number of white rows in which the total count is equal to or less than the threshold is determined to perform a second region cutout.

Further, in the second area cutout, a third area cutout is performed by the same processing as the first area cutout, and the same processing as the second area cutout is performed in the third area cutout. Performs the fourth region cutout.
Therefore, for the area detected by the fourth area cutout, attributes such as a character area, a photograph area, and a chart area are determined based on the run length information and the black pixel ratio information of the area.

[0005]

However, the conventional attribute discriminating method has the following problems.

(A) In the conventional attribute discriminating method, it is necessary to arbitrarily set various thresholds in accordance with a target in the region cutout processing. For example, when region segmentation is performed for individual journals, those thresholds must be reset to appropriate values according to each journal, and it takes time to select such values. Therefore, there is a problem that the efficiency of the attribute discrimination process performed after the region clipping process is low.

(B) To solve the problem (a),
The applicant of the present application has previously proposed an attribute discrimination method in Japanese Patent Application Laid-Open No. 1-264649. In this proposal, optimal 2
The document image is input by the binarization processing, the character / field separator area is subjected to image area separation, and the figure / table / photograph area is subjected to dither processing (halftone processing), and these two images are combined.
After scanning the image binarized by the optimal binarization processing to extract an area, the binarized image determined as a chart / photo candidate is scanned with a binarized image, and the image is scanned according to the number of white / black inversions. An area and a chart area are determined.

However, this attribute discrimination method has a problem that it is necessary to scan a dither-processed image in a vertical direction and a horizontal direction, which takes a long processing time. That is,
In the attribute discrimination method proposed earlier, for example, 400 dpi
(3408 x 4624 pixels) image is processed, and the entire image inside the rectangle obtained by the region extraction must be completely scanned once each in the vertical and horizontal directions, and this processing alone is close to one minute. Needs time. Therefore, considering the whole area division, it takes several minutes to process one document image, and the processing takes too much time.

As described above, it is difficult to determine the attributes of a photograph area and a chart area relatively easily in a short processing time, and it is difficult to obtain an attribute determination method that is technically sufficiently satisfactory. .

The present invention provides an attribute discriminating method which solves the problem of the prior art that it is difficult to discriminate the attributes of a photograph area and a chart area relatively easily with a short processing time. It is.

[0011]

According to a first aspect of the present invention, a document image is binarized to generate a binary image, and an information area (for example, a black pixel area) of the binary image is created. )
In the attribute discriminating method for detecting the circumscribed rectangle by converting the circumscribed rectangle into circumscribed rectangles and discriminating the attributes of the photograph area and the chart area of the circumscribed rectangle, the following means are taken.

A reduced image is created by reducing the entirety of the binary image, and the first circumscribed rectangle of the photograph candidate area and the chart candidate area is obtained from the features of the circumscribed rectangle and the features of the binary image within the circumscribed rectangle. Is extracted. Then, a second circumscribed rectangle corresponding to the reduced image is obtained from the first circumscribed rectangle, and the inside of the second circumscribed rectangle is scanned to count the number of all black pixels in the rectangle. By comparing the ratio of the counted total number of black pixels to the area of the second circumscribed rectangle with the threshold value, the photographic region and the chart region are separated.

According to a second aspect of the present invention, as in the first aspect, in the conventional attribute discriminating method, a photograph candidate area and a chart candidate area are determined from the characteristics of the circumscribed rectangle and the characteristics of the binary image in the circumscribed rectangle. A first circumscribed rectangle is extracted, and only the first circumscribed rectangle is reduced to create a reduced image. Then, a second circumscribing rectangle corresponding to the reduced image is obtained, the inside of the second circumscribing rectangle is scanned, and the number of all black pixels in the rectangle is counted. By comparing the percentage of the area of the circumscribed rectangle with the threshold value,
The photograph area and the chart area are separated.

According to a third aspect, in the first or second aspect, as a method for creating the binary image, the document image is read in the form of multi-valued image data, and the multi-valued image data is read with its characteristics and The optimal binarization is performed without deteriorating the properties.

According to a fourth aspect of the present invention, in the first or second aspect, the document image is obtained by binarizing the document image by simple binarization or dither processing alone. I have.

In a fifth aspect based on the first or second aspect, the reduced image is created by taking a logical sum (OR) between pixels.

[0017]

According to the first aspect of the present invention, since the attribute determining method is configured as described above, a binary image is created from a document image.
Further, when a circumscribed rectangle is detected from the binary image,
The entire value image is reduced to create a reduced image. Furthermore, a first circumscribed rectangle including a photograph candidate area and a chart candidate area is extracted from the features of the circumscribed rectangle before reduction and the features of the binary image in the circumscribed rectangle. Then, a second circumscribed rectangle corresponding to the reduced image is obtained from the first circumscribed rectangle,
By scanning the second circumscribed rectangle, the number of all black pixels in the rectangle is counted, and the photograph area and the chart area are separated by the ratio of the total number of black pixels to the area of the second circumscribed rectangle. Thereby, scanning of the reduced image is sufficient,
Further, since the number of times of the entire surface scan may be one, the processing can be simplified and the processing time can be shortened.

According to the second aspect, after extracting the first circumscribed rectangle including the photograph candidate area and the chart candidate area, only the first circumscribed rectangle is reduced to create a reduced image. Then, a second circumscribed rectangle corresponding to the reduced image is obtained, and the number of all black pixels in the second circumscribed rectangle is counted, as in the first invention. The photograph area and the chart area are separated according to the proportion of the rectangular area. By limiting the reduced area of the image to the chart / photograph area as described above, the processing speed can be further increased. According to the third aspect, the attribute is determined using the optimally binarized binary image data, so that the attribute can be accurately determined.

According to the fourth aspect, attribute determination is performed using binary image data binarized only by simple binarization or dither processing. Therefore, the binarization processing can be simplified.
By simplifying the subsequent processing, the speed of the attribute determination processing can be further increased.

According to the fifth aspect of the present invention, a reduced image is created by ORing image data such as pseudo halftone and halftone dot representation, so that most of the area is occupied by black pixels. It can be simplified.

[0021]

FIG. 2 is a schematic functional block diagram of an attribute discriminating apparatus showing a first embodiment of the present invention.

In this attribute discriminating apparatus, a CCD (Char)
An image sensor such as a Ge Coupled Device) is provided, and a reading unit 1 that reads a multi-value image data by scanning a document image with the image sensor is provided. The output side of the reading unit 1 is connected to a multi-valued image memory 2 for storing the read multi-valued image data, and an optimum binarizing unit 3. The optimal binarizing means 3 has a function of converting the read multi-valued image data into a binary image without impairing its characteristics and properties, and stores the converted binary image on its output side. The binary image memory 4 and the circumscribed rectangle detecting means 5 are connected.

The circumscribed rectangle detecting means 5 has a function of circumscribing and extracting an information area (for example, a black pixel area) of a binary image obtained by optimal binarization. A circumscribed rectangle position memory 6 for storing the position of the circumscribed rectangle thus determined, an attribute discriminating means 7, and a reduced binary image memory 8 are connected.

The attribute discriminating means 7 reduces the entire area of the extracted binary image data of the circumscribed rectangle to create a reduced image, stores the reduced image in the minimum binary image memory 8, and
From the extracted circumscribed rectangle and the characteristics of the reduced image, etc., the circumscribed rectangle is defined as a field separator area, a photograph area, a chart area,
And another character area. On the output side of the attribute discriminating means 7, there are provided a field separator position memory 9 for storing the positions of the classified field separator regions, a photograph region position memory 10 for storing the positions of the photograph regions, and a chart region for storing the positions of the chart regions. The position memory 11 and the character integrating means 12 are connected.

The character integrating means 12 has a function of integrating a line and a character block with respect to a circumscribed rectangle determined to be a character, and extracting a result of the integration as a character area. Is connected to a character area position memory 13 for storing the.

Next, an attribute determining method using the apparatus shown in FIG. 2 will be described.

When a document image is input to the reading unit 1, the reading unit 1 scans the input document image by an internal image sensor, reads multi-value image data, and stores the multi-value image data in a multi-value image memory. 2 is stored. The optimal binarizing means 3 inputs the multi-valued image data stored in the multi-valued image memory 2, performs image area separation on the character / field separator area, performs dither processing on the chart / photograph area, and processes both images. Perform processing for combining.

That is, the optimal binarizing means 3 determines the non-halftone image portion and the halftone image portion in the input multi-valued image data from the magnitude of the absolute value of the density difference value in the local region centering on the target pixel. Classify into parts and If the pixel is determined to be a non-halftone image portion, the pixel of interest is binarized using a fixed threshold. If the pixel is determined to be a halftone image portion, the pixel of interest is binarized using a dither matrix threshold.

When such an optimal binarization process is performed,
Non-halftone image parts such as character areas and chart areas become sharp images at the edges of the boundary between white and black, and halftone image parts such as photographic areas become dither images seen in newspapers and the like. Obtained as a binary image. The data of the composite binary image is stored in the binary image memory 4.

The circumscribed rectangle detecting means 5 scans the binary image data stored in the binary image memory 4 in, for example, the main scanning direction (horizontal direction) X and the sub-scanning direction (vertical direction) Y.
Then, in all areas of the binary image data, a black line in which at least one black pixel exists in the horizontal direction X is detected, and an area until a white line in which no black pixel exists continues more than the threshold TH is determined. Cut out (step ST1). Next, a black column in which at least one black pixel exists in the vertical direction Y is detected in the cut-out region, and a region is cut out until a white column in which no black pixel exists continues beyond the threshold value TH (step ST2).

Further, in the cut area,
In the same manner as described above, the steps ST1 and ST2 are repeatedly executed, and the repetition processing is continued until further cutout cannot be performed. Then, the finally cut out area is detected as a circumscribed rectangle, and the start point coordinates and the end point coordinates indicating the position are stored in the circumscribed rectangle position memory 6.

The attribute discriminating means 7 reduces the binary image stored in the binary image memory 4 to form a reduced image, and stores the reduced image in the reduced binary image memory 8. further,
All the circumscribed rectangles stored in the circumscribed rectangle position memory 6 are classified into three attributes of a field separator area, a photograph / graphic area, and other character areas from the circumscribed rectangle features and the features within the rectangle. Further, referring to the reduced image stored in the minimum binary image memory 8, a photograph area and a chart area are determined from the photograph and chart area.

The start point coordinates and the end point coordinates indicating the positions of the field separator area, the photograph area, and the graphic area classified by the attribute discriminating means 7 are stored in the field separator position memory 9, the photograph area position memory 10, and the chart area position, respectively. Stored in the memory 11. Further, with regard to the character area, the circumscribed rectangle determined as the character area is integrated into a line / character block by the character integration means 12, and the result is extracted as a character area and the character area position memory 1 is extracted.
3 is stored.

FIG. 1 is a flowchart of the attribute discriminating process performed by the attribute discriminating means 7 in FIG.

In this attribute discriminating process, in step S20, the binary image data stored in the binary image memory 4 of FIG. 2 is input, and in step S21, the binary image data is reduced in size. When the reduction ratio is realized by a program, for example, 1/8 (= 400 DPI image)
50 DPI) is appropriate. Next, step S22
, The start point coordinates (XSi, YSi) and the end point coordinates (XEi, YEi) of the circumscribed rectangle of the area whose attribute is to be determined and stored in the binary image memory 4 are input. Here, X is the main scanning direction (horizontal direction), Y is the sub-scanning direction (vertical direction), and the coordinate number i is i = 1, 2,..., N.

In step S23, the input circumscribed rectangle has a vertical HI (= YEi-YSi + 1) and a horizontal WD (= XE
It is compared whether the aspect ratio WD / HI with (i-XSi + 1) is larger than a fixed threshold TH1. If the aspect ratio WD / HI of the circumscribed rectangle is larger than the fixed threshold value TH1, the circumscribed rectangle is determined as the field separator area in step S24, and the determination result is stored in the field separator position memory 9 in FIG.

The aspect ratio WD / HI of the circumscribed rectangle is a fixed threshold T
If it is smaller than H1, it is set as a character / photo / diagram candidate area. For this character / photo / diagram candidate area, in step S25, the area WD × HI of the circumscribed rectangular area is obtained, and it is compared whether the area WD × HI is larger than a certain fixed threshold TH2. If it is larger, it is determined to be a photograph / chart candidate area, and the process proceeds to step S26. If it is smaller, it is determined in step S31 that the area is a character area.
Through the character area position memory 13.

If it is determined in step S25 that the area is a photograph / chart candidate area, the process proceeds to step S26 where step S26 is performed.
The start point coordinates (XSi / n, YSi / n) and end point coordinates (XEi /
After calculating (n, YEi / n), in step S27, the reduced image is scanned in the horizontal direction X and the vertical direction Y within the circumscribed rectangle, and the total number of black pixels TO occupying the inside of the circumscribed rectangle is counted. Then, the ratio of the total number of black pixels TO to the area of the circumscribed rectangle is calculated, and it is compared whether or not the ratio is larger than a certain fixed threshold TH3.

If the ratio is greater than the fixed threshold TH3,
In step S29, the circumscribed rectangular area is determined to be a photograph area,
The determination result is stored in the photograph area position memory 10 in FIG. If the ratio is smaller than the fixed threshold TH3, in step S30, the circumscribed rectangular area is determined to be a chart area, and the determination result is stored in the chart area position memory 11 of FIG. Thus, the attribute determination processing ends.

As described above, in the first embodiment, the entire binary image data is reduced in step S21 in FIG.
In step S27, a photograph area and a chart area are determined by processing such as scanning a reduced image.
Therefore, the scanning process is performed on the reduced image, and the number of times the entire surface needs to be scanned is only one. Therefore, the discrimination between the photograph area and the chart area is simplified, and the processing time for discriminating the attribute is also increased. Can be greatly reduced, and high-speed processing can be performed.

FIG. 3 is a flowchart of an attribute discriminating process according to a second embodiment of the present invention. Elements common to those in the flowchart of FIG. 1 are denoted by the same reference numerals.

In the flowchart of the attribute determination process,
The processing content of the area determination when the reduced area of the image is changed is shown.

That is, in this attribute discriminating process, step S21 in FIG. 1 is omitted, binary image data is input in step S20, and a circumscribed rectangle is input in step S22 without reducing the binary image data. Then, in step S23, the field separator area and the character / photograph / chart candidate area are classified.

Then, in step S25, characters, photos,
The area of the circumscribed rectangular area is determined for the chart candidate area, and the photograph / chart candidate area and the character area are classified. For the photograph / chart candidate area, the internal binary image at the start point coordinates (XSi, YSi) and end point coordinates (XEi, YEi) of the circumscribed rectangle is reduced by 1 / n. Next, in step S26, as in FIG. 1, the start point coordinates (XSi / n, YSi / n) and the end point coordinates (XEi / n, YEi / n) of the circumscribed rectangle corresponding to the reduced image are calculated. S
Through the steps 27 to S30, the photograph area and the chart area are determined.

In the second embodiment, the reduced image is used only when discriminating between the photograph area and the chart area, and is not used when discriminating the field separator area and the character area. It is limited to the chart / photo candidate area. As a result, the amount of reduction processing can be reduced as compared with the first embodiment, and higher-speed attribute discrimination processing can be performed.

The present invention is not limited to the above embodiment,
Various modifications are possible. For example, there are the following modifications.

(I) FIGS. 4 (a) and 4 (b) are diagrams showing an example of the relationship between the binarization processing and the reduced image, and FIG. 4 (a) shows the binary image data in pseudo halftone expression. FIG. 3B is a diagram showing a reduced image when the binary image data is a halftone dot.

As is apparent from this figure, the photographic area is reduced by taking the OR between the pixels, so that even if the binarization processing is a pseudo-halftone / halftone expression, almost all areas are black pixels. Will be occupied by. Therefore, in the above embodiment, the multi-valued image data is optimally binarized by the optimal binarizing means 3 in FIG.
Although the image data is binarized, it is possible to discriminate between a photograph area and a chart area by using a simple binary image or an image binarized only by dither processing. As described above, by using an image binarized by simple binary or dither processing alone, the processing speed can be further increased as compared with the above embodiment. (Ii) The attribute discriminating apparatus of FIG. 2 that executes the attribute discriminating methods of the first and second embodiments may be changed to another configuration. For example, the positions of the circumscribed rectangles determined as the field separator area, the photograph area, and the chart area are stored in the memories 9, 10, and 11, respectively, and the circumscribed rectangle positions of the other character areas are stored in the circumscribed rectangle position memory 6 as they are. Has been saved. Therefore, the circumscribed rectangular position memory 6 can serve as the character area position memory 13, and thus, by omitting the character area position memory 13,
Memory can be reduced.

[0049]

As described in detail above, according to the first aspect, a reduced image is created by reducing the whole of a binary image, and the reduced image is scanned to define a circumscribed rectangle in the reduced image. The photograph area and the chart area are determined based on the ratio of the total number of black pixels to the image area. Therefore, since the reduced image is scanned and the entire surface is scanned only once, the photograph area and the chart area can be easily distinguished from each other, and the processing time for attribute determination can be greatly reduced.

According to the second aspect of the present invention, only the first circumscribed rectangle is reduced only in the chart / photograph area,
Since a reduced image is created, the processing involved in the reduction processing and the like can be simplified, and the speed of the discrimination processing can be further increased.

According to the third aspect, since the attribute is determined using the optimal binarized image, it is possible to accurately determine the photographic area and the chart area.

According to the fourth aspect, attribute determination is performed using a binary image binarized by simple binary or dither processing alone, so that the determination processing is simplified and speeded up. Can be planned.

According to the fifth aspect, since the reduced image is created by ORing the pixels, the creation of the reduced image is simplified, and the attribute discrimination process can be further speeded up.

[Brief description of the drawings]

FIG. 1 is a flowchart of an attribute determination process according to a first embodiment of the present invention.

FIG. 2 is a functional block diagram of the attribute discriminating apparatus according to the first embodiment of the present invention.

FIG. 3 is a flowchart of an attribute determining process according to a second embodiment of the present invention.

FIG. 4 is a diagram illustrating an example of a relationship between a binarization process in FIG. 2 and a reduced image thereof.

[Explanation of symbols]

 REFERENCE SIGNS LIST 1 reading unit 3 optimal binarizing means 5 circumscribed rectangle detecting means 7 attribute discriminating means 8 reduced binary image memory

──────────────────────────────────────────────────続 き Continuation of front page (56) References JP-A-3-126180 (JP, A) JP-A-1-316883 (JP, A) JP-A-56-116182 (JP, A) (58) Survey Field (Int.Cl. ⁷ , DB name) H04N 1/40-1/409 G06T 1/00

Claims (5)

(57) [Claims] 1. A binary image is created by binarizing a document image, a circumscribed rectangle is formed in an information area of the binary image, a circumscribed rectangle is detected, and attributes of the photograph area and the chart area of the circumscribed rectangle are determined. In the attribute discriminating method, a reduced image is created by reducing the entirety of the binary image, and the first image of the photograph candidate area and the chart candidate area is determined from the features of the circumscribed rectangle and the features of the binary image within the circumscribed rectangle. From the first circumscribed rectangle, obtain a second circumscribed rectangle corresponding to the reduced image from the first circumscribed rectangle, scan the inside of the second circumscribed rectangle, and scan all the circumscribed rectangles within the second circumscribed rectangle. An attribute discriminating method comprising: counting the number of black pixels; and comparing a ratio of the total number of black pixels to the area of the second circumscribed rectangle with a threshold value to separate a photograph region and a chart region. 2. An attribute discriminating method for binarizing a document image to create a binary image, circumscribing an information area of the binary image to detect a circumscribed rectangle, and discriminating an attribute of the circumscribed rectangle. Based on the features of the circumscribed rectangle and the features of the binary image within the circumscribed rectangle, the first of the photograph candidate area and the chart candidate area
Is extracted, a reduced image is created by reducing only the first circumscribed rectangle, a second circumscribed rectangle corresponding to the reduced image is obtained, and the inside of the second circumscribed rectangle is scanned. Counting the number of all black pixels in the second circumscribed rectangle, and comparing the ratio of the total number of black pixels to the area of the second circumscribed rectangle with a threshold value to separate the photographic region and the chart region. An attribute determination method characterized by the following. 3. The attribute discriminating method according to claim 1, wherein the binary image is obtained by reading the document image in the form of multi-valued image data, and losing the multi-valued image data without losing its characteristics and properties. Attribute discrimination method obtained by optimal binarization. 4. The attribute discriminating method according to claim 1, wherein the binary image is obtained by binarizing the document image by simple binarization or dither processing alone. 5. The attribute determining method according to claim 1, wherein the reduced image is created by taking a logical sum between pixels.