JP3488678B2 - Image classification device - Google Patents

Description

Detailed Description of the Invention

[0001]

BACKGROUND OF THE INVENTION 1. Field of the Invention The present invention relates to an image classifying apparatus, and more particularly to an image classifying apparatus for automatically classifying and registering attributes of input images when inputting an image database.

[0002]

2. Description of the Related Art A method of automatically dividing the inside of an image into which an image is determined to be a document image by a human and performing attribute determination of each region is disclosed in Japanese Patent Laid-Open No. 4-316.
180 and (B) Japanese Patent Laid-Open No. 62-71379.

The former (a) divides a binarized input image into regions, and determines attributes such as a photo region or a character region based on the number of black and white inversions in each region and the ratio of black pixels. (B) similarly divides the binarized input image into regions, and determines attributes such as a photo region or a character region based on the run-length characteristics and the black pixel ratio of each region.

[0004]

However, the conventional technique only determines the attribute of each area divided into areas in the document image, and determines whether the input image itself is the document image. The method and apparatus for determining whether or not has not been disclosed so far.

Therefore, when an image is registered in a multimedia database into which various image data are input, not limited to a document image, a person classifies and judges the contents of the input image and adds an appropriate keyword. There was a need.

Further, in such a multimedia database, Japanese Patent Application Laid-Open No. 4-316180 and Japanese Patent Application Laid-Open No. 62-216180 are disclosed.
When applying the techniques of JP-A-71379, it is necessary for a person to determine in advance whether or not the input image is a document image before applying the techniques.

Further, since there is no conventional apparatus for classifying only the business card images from the image data, when performing processing such as address book creation from the business card images by utilizing, for example, character recognition, a human is selected from the various images. Had to be selected and the character recognition processing was performed on each image.

The present invention has been made in order to solve the above problems, and automatically determines whether the input image itself is a document image, particularly whether it is a business card image or not. An image that can be classified (adding keywords) and registered in the image database without the trouble of, and that can automatically perform processing according to the content of the image, such as address book creation using character recognition. An object is to provide a classification device.

[0009]

An image classification apparatus according to the present invention is configured to identify, from input image data, a connected component which is a plurality of adjacent pixels and which represents an object, and a number of connected components. Measuring means for measuring as a feature amount,
And a determination unit that determines whether or not the input image data is a business card image based on the measured feature amount.

The image classifying apparatus according to the present invention comprises an identifying means for identifying a connected component consisting of a plurality of adjacent pixels representing an object from the input image data, and two adjacently adjacent identified connected components. Direction extracting means for dividing each connected component and extracting the arrangement direction for each of the two connected components; measuring means for measuring the degree of concentration of the extracted arrangement direction in the direction of a specific range as a feature amount; And a determination unit that determines whether or not the input image data is a business card image based on the measured feature amount.

The image classification apparatus according to the present invention comprises: differential image data creating means for creating differential image data from input image data; density projection histogram creating means for creating a density projection histogram based on the differential image data; It is provided with a calculating means for calculating a feature amount from the created density projection histogram, and a determining means for determining whether or not the input image data is a business card image based on the calculated feature amount.

The image classifying apparatus according to the present invention is the above-described image classifying apparatus, in which the identified connected component is divided into two most adjacent connected components, and the divided two connected components are arranged. A direction extracting means for extracting the direction and a measuring means for measuring the degree of concentration of the extracted arrangement direction in the direction of a specific range as a feature amount are provided, and the feature amount indicates the degree of concentration in the direction of the specific range. It includes.

The image classifying apparatus according to the present invention is the above-mentioned image classifying apparatus, wherein differential image data creating means for creating differential image data from input image data, and density based on the differential image data. The present invention further comprises density projection histogram creating means for creating a projection histogram, and calculation means for calculating data based on the differential value of the created density projection histogram, and the feature amount includes data based on the differential value of the density projection histogram.

The image classifying apparatus according to the present invention is the above-mentioned image classifying apparatus, wherein the identifying means sets a predetermined condition among a plurality of adjacent images representing an object from the input image data. Pixels to be filled are identified as connected components.

(Operation) The image classification device according to the present invention identifies a connected component consisting of a plurality of adjacent pixels representing an object from the input image data, and measures the number of connected components as a feature quantity. Based on the measured feature amount, it is determined whether the input image data is a business card image.

The image classification apparatus according to the present invention identifies a connected component consisting of a plurality of adjacent pixels representing an object from the input image data, and identifies the identified connected component for every two adjacent connected components. The arrangement direction of each of the two connected components that have been separated is extracted, the degree of concentration of the extracted arrangement direction in a specific range is measured as a characteristic amount, and the image data input based on the characteristic amount is extracted. Is a business card image.

The image classification apparatus according to the present invention creates differential image data from the input image data, creates a density projection histogram based on the differential image data, and calculates a feature amount from the created density projection histogram. It is determined whether the input image data is a business card image based on the feature amount.

In addition to the function of the above-mentioned image classifying device, the image classifying device according to the present invention divides the identified connected component into two most adjacent connected components and extracts the arrangement direction of each of the two connected components. Then, the degree of concentration of the extracted arrangement direction in a specific range is measured as a feature amount.

In addition to the operation of the image classification device, the image classification device according to the present invention creates differential image data from the input image data, creates a density projection histogram based on the differential image data, and creates the created density. Data based on the differential value of the projection histogram is used as the feature amount.

In addition to the operation of the above-described image classification device, the image classification device according to the present invention identifies, from the input image data, a pixel satisfying a predetermined condition among a plurality of adjacent pixels representing an object as a connected component. To do.

[0021]

BEST MODE FOR CARRYING OUT THE INVENTION Embodiments of an image classification apparatus according to the present invention will be described below in detail with reference to the drawings.

(First Embodiment) FIG. 1 is a block diagram showing the system configuration of an image classification apparatus according to the first embodiment of the present invention.

Referring to the figure, the image classifying apparatus roughly includes an image input apparatus 11 for inputting an image, which includes a camera, a scanner, etc., and processes the input image into a document image and a non-document image. The image processing device 12 for classifying (adding keywords), and the image database device 13 for storing the images classified by the image processing device 12.

The respective devices are connected by a bus 15. The image processing device 12 includes a control device 121 configured by a CPU or the like for processing and classifying an input image and controlling the operation of the entire system, a ROM 122 for storing various programs and constants, an image and A RAM 123 for storing processing data and the like is included.

The ROM 122 includes an image input processing unit 1221 used for image input processing, an image binarization processing unit 1222 used for binarizing an image, and binary image data. Area extraction processing section 1223 for identifying the area of the image, a labeling processing section 1224 for performing labeling (processing for assigning a label number) to the connected components of the data in the binarized image data, and the binarized image data Character-likeness feature extraction processing unit 1225 for extracting the character-likeness features in the image, image classification processing unit 1226 for classifying images, and image registration processing unit 1227 for classifying and registering image data in the image database device 13. Including and

The RAM 123 is a multi-valued image memory 1231 that stores the input image data as image data of 256 gradations, and a binary image memory that stores the image data represented by the binary values "0" and "1". 1232, a label image memory 1233 that stores the label number of each connected component of the data, a fillet coordinate memory 1234 that stores the fillet coordinates and area of each connected component, and a differential image of the image data of the multivalued image memory 1231. A differential image memory 1235, a differential binary image memory 1236 that stores image data obtained by binarizing the image data of the differential image memory 1235 with a threshold value, and a binary image memory 123.
AND for storing a logical product image (AND image) of the image data of 2 and the image data of the differential binary image memory 1236
And an image memory 1237.

With this system configuration, the image input device 1
Various images input from the image processing device 12
By checking the characteristic amount of the image in, the image is automatically classified into the document image or the non-document image and recorded in the image database device 13.

For example, the document image is an image containing many characters shown in FIG. 4, and the non-document image is the landscape image shown in FIG.

2 and 3 show the controller 121 of FIG.
3 is a flowchart showing the processing routine of FIG.

The flowcharts of FIGS. 2 and 3 are connected at the portion A in the figures and show a series of processes.

In step S201, the controller 12
1 inputs image data by the image input device 11 and stores it in the multi-valued image memory 1231. The image data stored in the multi-valued image memory 1231 is X
The data consists of 512 pixels in the direction and 432 pixels in the Y direction, and one pixel has a brightness value of 256 gradations (8 bits) from “0” (black pixel) to “255”. The upper left pixel in the image data diagram is the origin (0,0), and X
The coordinates of 0 to 511 are attached to the direction, and the coordinates of 0 to 431 are attached to the Y direction.

The multi-valued image memory 1231 has the structure shown in FIG. That is, the multivalued image memory 1231 has X
It has an 8-bit memory for storing the brightness value of 256 gradations for each pixel indicated by the coordinate and the Y coordinate.
In the memory, the luminance value of the pixel at the origin (0, 0) is stored at the beginning, and the data of the luminance value of the pixel is sequentially stored in the increasing order of the X coordinate. The brightness value data is stored in the order of increasing coordinates, and the coordinates (5
The data of the luminance values of all the pixels up to 11, 431) are stored.

In step S202, the controller 12
1 is the entire image data (multi-valued image data) stored in the multi-valued image memory 1231 and having a brightness value of 256 gradations.
Quantify. Binary image data (binary image data)
Are stored in the binary image memory 1232.

In the binarization, specifically, after the control device 121 scans the entire multi-valued image data in the memory 1231 and sets the threshold value of the brightness for the binarization, the brightness equal to or higher than the threshold value is set. This is performed by setting a pixel having a pixel as a white pixel (data of "1") and a pixel having a luminance less than the threshold value as a black pixel (data of "0"). As a threshold setting method, for example, supervision by Tamura Hideyuki, edited by Japan Industrial Technology Center,
"Introduction to Computer Image Processing", Soken Shuppan, 1988
Year, p. 66-p. 69 (discriminant analysis method, p-tile method, modal method, etc.) can be used. In this example, the discriminant analysis method therein is used.

In the discriminant analysis method, when it is assumed that a set of pixels having different brightness values is divided into two classes (greater than or equal to th and less than th) by a threshold value (th), the variance between the respective classes is maximum. The threshold value (th) is set so that

After the threshold value is determined, the control device 121 scans the multi-valued image data stored in the multi-valued image memory 1231 from the origin, reads the brightness value of each pixel, and compares it with the set threshold value. To do. As a result, if the brightness value is less than the threshold value, the brightness value of the pixel is “0” (black pixel), and if the brightness value is more than the threshold value, the brightness value of the pixel is “1” ( Binary image memory 123 as white pixels)
Stored in 2.

The address configuration of the binary image memory 1232 is the same as that of the multi-valued image memory 1231 of FIG. 7, except that 1-bit data of "0" or "1" is recorded as the brightness value. .

In step S203, the controller 12
1 determines whether the number of white pixels in the binary image data of the binary image memory 1232 is greater than 1/3 of the total number of pixels. Y
If it is ES, in step S204, "0" data and "1" of the binary image data of the binary image memory 1232 are obtained.
The data of is inverted.

If NO at step S203, the process of step S204 is not performed.

The processing of steps S203 to S204 is performed for the following reason. Since image data is normally processed with black pixels as the background and white pixels as the object, for example, when image data of an object with black pixels is input to the background of white pixels shown in FIG. It is necessary to invert black and white. Therefore, when the number of white pixels is ⅓ or more in all the pixels, it is determined that the background is a white pixel and the object is a black pixel, and the white pixel and the black pixel are inverted.

In step S205, the connected components of the white pixels in the binary image data of the binary image memory 1232 are extracted, and labeling is performed for each.

The connected component refers to a group of white pixels (data of "1") which are continuous vertically, horizontally and diagonally in the image data. For example, if the image data is the data shown in FIG. 8, four connected components of connected components 20a to 20d are extracted as a group in which data of "1" is continuous vertically and horizontally, and each connected component is identified. The label number for this is set as shown in FIG. In FIG. 9, as the label numbers, the connected component 20a in FIG. 8 is "1" data, the connected component 20b is "2" data, and the connected component 20c is "3".
And the data of "4" are set in the connected component 20d. For a black pixel having no label number, "0" data is set instead of the label number.

The label number is the label image memory 1233.
Recorded in. The label image memory 1233 has the address configuration shown in FIG. That is, the address configuration is the same as that of the binary image memory 1232. Each address has a 16-bit memory for storing the label number attached to the image corresponding to the address. It is necessary to prepare the number of bits of each memory so that the maximum value of the label number can be stored, but in the present embodiment, 16 bits is considered to be sufficient.

As a labeling method, for example, "Digital Image Processing for Image Understanding (I.
I) ”Shokoido, 1988, p. 45-p. The method described in 46 can be used.

In step S206, the controller 12
1 extracts the fillet coordinates and area of each connected component with a label number. Fillet coordinates are the upper left and lower right coordinates of the smallest rectangle circumscribing one connected component in the image. When there are connected components as shown in FIG. 11, the upper left fillet coordinates are (x_sp, y_sp ), And the fillet coordinate at the lower right is (x_ep, y_ep).

That is, the maximum X coordinate and the maximum Y coordinate among the X coordinate and Y coordinate of the pixels having the same label number.
The set of coordinates is the lower right fillet coordinate, and the set of the minimum X coordinate and the minimum Y coordinate is the upper left fillet coordinate.

The area of the connected component is represented by the number of pixels having the same label number, that is, the number of pixels of each connected component.

In FIG. 11, the connected components are shown in black and the background is shown in white for convenience of explanation. The fillet coordinates and area of each connected component are stored in the fillet coordinate memory 1234. The fillet coordinate memory 1234 has the configuration shown in FIG. That is, the X coordinate and the Y coordinate of the upper left fillet coordinate, the X coordinate and the Y coordinate of the lower right fillet coordinate, and the area of the connected component are recorded in the order of label numbers. The memory for storing each of the coordinates and the area is 16 bits. Therefore, each label has a memory area of 80 bits.

In step S207, unnecessary connected components are removed. The unnecessary connected component is the connected component described below.

(1) Minute component in which the area of the connected component is 3 pixels or less (2) The width in the X direction of the rectangle circumscribing the connected component is 1/8 of the width (512 pixels) of the entire image data in the X direction. A huge component whose width in the Y direction is greater than or equal to one-eighth of the width of the entire image data in the Y direction (432 pixels) (3) A rectangle circumscribing the connected component in the X direction Slender component in which one of the widths in the Y direction is greater than four times the width of the other is considered to be not the character information and thus deleted from the binary image memory 1232.

For example, in the binary image memory 1232, as shown in FIG.
There are six types of connected components, labeled 1 to 6 shown in FIG. 3, and the fillet coordinate memory 12
It is assumed that the data shown in FIG. 14 is recorded in 34. At this time, since the connected component of the label 2 has a too large width in the Y direction, it is removed as the above-mentioned huge component.

The connected component of the label 4 is removed as a minute component because its area is 3 pixels or less.

The connected component of label 6 has a width in the X-axis direction of Y.
Since it is four times or more the width in the axial direction, it is removed as an elongated component.

Therefore, as a result of the removal of the respective connected components, as shown in FIG.
, The connected components of labels 1, 3, and 5 are left.

In step S208, the controller 12
1 performs spatial differentiation processing on the multi-valued image data in the multi-valued image memory 1231. The image subjected to the differential processing (differential image) is recorded in the differential image memory 1235.

Differentiation of an image is carried out to detect a portion having a high density discontinuity in the image.

A portion having a high density discontinuity in the image (where the change in the density value is large) is recognized as a boundary (edge portion) of the object.

A specific method of differentiating processing is performed as follows. Referring to FIG. 16, a multi-valued image memory 1231
If the coordinate of the image to be differentiated in is the coordinate (i, j), the pixel of the coordinate (i, j) and the pixel of the coordinate (i, j) are vertically and horizontally diagonally touched, and the total length is 3 × width 3 And the pixel of is the pixel used for differentiation. The luminance value of each of the vertical 3 × horizontal 3 pixels is multiplied by the coefficient shown in FIG. _Let δ _x F (i, j) be the sum of all the vertical 3 × horizontal 3 luminance values after being multiplied by the coefficient.

On the other hand, a value obtained by multiplying each luminance value of vertical 3 × horizontal 3 pixels by the coefficient shown in FIG. 18 and adding all the luminance values of vertical 3 × horizontal 3 after multiplying the coefficient. Δ
_{Let y} F (i, j).

Here, as the differential value, for example, | δ _x F
The value obtained by (i, j) | + | δ _y F (i, j) | can be used.

The coefficients in FIGS. 17 and 18 are called Sobel operators and are supervised by Hideyuki Tamura, edited by Japan Industrial Technology Center, "Introduction to Computer Image Processing", Soken Shuppan, 198.
8 years, p. 118-p. At 125. Other methods may be used as the method of obtaining the differential value.

The pixels in the multi-valued image memory 1231 are scanned by the Sobel operator in FIGS. 17 and 18, and the differential value of each pixel is recorded in the differential image memory 1235.

In the multi-valued image memory 1231, the Sobel operator cannot be applied to the portion corresponding to the edge of the image data (X coordinate is 0 or 511 or Y coordinate is 0 or 431). Data of "0" is recorded.

In step S209, the differential image recorded in the differential image memory 1235 is binarized by a threshold value and recorded in the differential binary image memory 1236. Two
Since the value conversion process is the same as the method described in step S202, the description thereof is omitted here.

By these processes, the differential binary image memory 1
In 236, a portion of the input image where the change in the luminance value is large is recorded as a white pixel (data of “1”).

In step S210, the image data in the binary image memory 1232 and the differential binary image memory 1236 are stored.
AND operation of the image data of
It is recorded in the D image memory 1237. What is AND operation 2
Data of the same coordinate position of two binary images are both "1"
This is an operation in which "1" data is given only to pixels which are (white pixels) and "0" data is given to the other pixels.
By this processing, the overlapping area of the binary image data of the effective connected component and the binary image data of the portion where the change of the luminance value is large is recorded in the AND image memory 1237 as the data of “1”.

FIG. 19 is a diagram showing a state where one connected component in the document image of FIG. 4 is recorded in the binary image memory 1232.

Note that, in FIG. 19, white pixels are used as a background (data portion of "0") and black pixels are objects (data portion of "1") for convenience of description.
The black pixel is used as the background (that is, the black and white is inverted from that in FIG. 19)
It is recorded in the memory (the same applies to FIG. 24 below).

FIG. 20 is a diagram showing a state in which the same connected component as the connected component shown in FIG. 19 is recorded in the differential binary image memory 1234. Referring to the figure, the boundary between the object having a large change in luminance value and the background is recorded as "1" data.

FIG. 21 is a diagram for explaining the same connected component as that shown in FIG. 19 when it is recorded in the AND image memory 1237. The AND image memory 1237 stores the result of the AND operation of the image of FIG. 19 and the image of FIG.

22 to 24 are diagrams showing a state in which one connected component in the landscape image of FIG. 5 is recorded in each memory, and are diagrams corresponding to FIGS. 19 to 21.

There is the following tendency when comparing the landscape image and the document image. In the document image, the AND image memory 12 for the area of the connected component recorded in the binary image memory 1232
The area ratio of the connected component stored in 37 is relatively large, and the ratio of the area of the connected component stored in the AND image memory 1237 to the area of the connected component stored in the binary image memory 1232 in the landscape image is Relatively small. This is based on the fact that the connected component of characters has a lot of edge information, and it is possible to determine whether the input image is a document image or other than that by the difference.

In step S211, the controller 12
1 calculates the ratio edge ratio of the total area of the AND image memory 1237 calculated in step S210 to the total area of the effective connected component calculated in step S207. The edge ratio is calculated by the equation (1).

Edge ratio = (Σedge area / Σbinary area) × 100 (1) In the expression (1), the edge area is the area of each region recorded in the AND image memory 1237, and bina
The ry area is the area of each region included in the binary image memory 1232. Σ indicates that all areas in the memory are summed.

The edge ratio is calculated in each area,
Finally, the average of the entire edge ratio may be taken.

In step S212, the controller 12
In step 1, the edge ratio obtained in step S211 is compared with the set threshold value, and it is determined whether the edge ratio is greater than or equal to the threshold value.

If YES in step S212, it is determined in step S213 that the image data is a document image.

If NO in step S212, it is determined in step S215 that the image data is non-document.

A value of 60%, for example, is used as the threshold value, but it may be changed according to the judgment of the user.

In step S214, the image is stored in the image database device 13 together with the determined keywords (attributes) of document or non-document.

FIG. 25 is a diagram for explaining the implementation results (including effects) in this embodiment. The horizontal axis is edge rat
io, the vertical axis represents the number of sample images classified into a predetermined range of edge ratio. The edge ratio is divided in a range of 5%, and the number of document and non-document sample images classified in the range is shown as a histogram.

Referring to the graph, the document image has an edge ratio
Is greater than 60%, and non-document images have an edge ratio of 60%
Tends to be smaller. From this, it is understood that the document image and the non-document image can be classified by using the value of 60% as the threshold value in the third step S212.

As described above, in the present embodiment, by utilizing the feature that the character area has a lot of edge information, the ratio of the area of the differential binary image (representing the edge information) in the effective connected image is examined. Thus, it is possible to accurately classify a document image including many character regions and a non-document image.

Therefore, such an image classifying device can be used as an image database or an input device of a facsimile machine. For example, when used as an input device for an image database, documents and non-documents are automatically classified, and the keywords "document" or "non-document" are automatically added to the input images without human intervention. It will be possible.

Further, by applying character recognition to an image classified as a document image, a keyword can be automatically extracted from the document in the image and added.

Further, by using the image classifying apparatus of the present invention when inputting the image database, the user can input the document image and the non-document image to one system without being aware, and at the time of the search, the document image and the non-document image can be input. It is possible to search individually.

Further, by using the present invention as an input device of a facsimile device, an input image is a document,
It becomes possible to classify the document into non-documents, and for example, it becomes possible to automatically select a printing mode.

Further, the image classification apparatus according to the present invention is used for attribute determination of each area by dividing an image in which a document, a photograph, a drawing and the like are mixed into each area and then applying the divided area to each area. Is possible.

(Second Embodiment) FIG. 26 shows a second embodiment of the present invention.
3 is a block diagram showing a device configuration of an image classification device in the embodiment of FIG.

Referring to the figure, this block diagram is obtained by removing the differential binary image memory 1236 and the AND image memory 1237 from the block diagram of FIG.

The basic processing performed in this embodiment is the same as that in the first embodiment, but the area of the circumscribed rectangle of the connected component and the inside thereof are used as a criterion for classifying the document image data and the non-document image data. It is characterized by using the ratio with the area of the effective connected component.

The processing in the image classifying apparatus in this embodiment is performed as follows. First, the processing from steps S201 to S207 in FIG. 2 is performed. Since this process is substantially the same as the process described in the first embodiment, the description thereof will not be repeated here.

After the processing of step S207 is performed,
The process of the flowchart shown in FIG. 27 is performed.

In step S401, the controller 12
1 calculates the ratio of the area of the effective connected component to the area of the circumscribed rectangle. The area of the effective connected component is read from the fillet coordinate memory 1234, and the area of the circumscribed rectangle is obtained from the fillet coordinate of the fillet coordinate memory 1234.

Area of circumscribed rectangle of effective connected component (filet
area) of the effective connected component (binary area)
The bin ratio of is calculated by the equation (2).

Bin ratio = (Σbinary area / Σfilet area) × 100 (2) In the equation (2), Σ indicates the sum of the entire image.

In this embodiment, the ratio of the total area of the effective connected components to the total area of the circumscribed rectangles of the effective connected components is used as the bin ratio. However, for each effective connected component A method may be used in which the ratio of the area of the effective connected component to the area of the circumscribed rectangle is calculated and then the average value of the entire area is taken.

In step S402, the controller 12
Bin rati obtained in step S401 by 1
It is determined whether o is less than a predetermined threshold value stored in advance. As the predetermined threshold value, for example, a value of 55% is used as described later, but the threshold value may be arbitrarily set according to the judgment of the user. At that time, the processing routine for setting the threshold value by the user is RO
It is recorded in M122.

If YES in step S402, the input image is determined to be a document image in step S403.

If NO in step S402, the input image is determined to be a non-document image in step S404.

In step S405, the image data is recorded in the image database device 13 together with the keywords (attributes) document / non-document.

FIG. 28 shows an example of binary image data in a document image, and FIG. 29 shows an example of binary image data in a landscape image (non-document image).

In the figure, the dotted line indicates the circumscribed rectangle in each connected component. For convenience of explanation, the object is shown in black and the background is shown in white in the figure.

Referring to the figure, the ratio of the area of the effective connected component to the area of the circumscribed rectangle is smaller in the document image than in the non-document image. Thereby, the document image and the non-document image can be classified.

FIG. 30 shows bi having a document image and a non-document image.
It is a figure for demonstrating n ratio. The abscissa represents the bin ratio divided by 5%, and the ordinate represents the number of sample images classified in each bin divided by 5%.

Referring to the figure, the document image has a bin ratio of 5
It is less than 5%, and the nonratio image has a bin ratio of 55% or more. Therefore, it is possible to classify the document image and the non-document image by setting the threshold value in step S402 of FIG. 27 to 55% and determining the document image below the threshold value and the non-document image above the threshold value. Recognize.

As described above, in this embodiment, the characteristic of the character region that the proportion of the connected components in the circumscribed rectangle is small because it is a line figure, and the effective connected component area of the circumscribed rectangle is By using the ratio, the feature of the document image in which many character areas are arranged side by side can be extracted well and classified accurately.

Also, ed used in the first and second embodiments
It is possible to classify image data with higher accuracy by combining and determining a plurality of feature quantities indicating character likeness including ge ratio and bin ratio.

For example, in one image data
A method is used to classify image data based on the total score, which is based on the score of the document image based on the edge ratio and the document image based on the bin ratio, or the image determined near the threshold value in one feature amount. The data can be realized by a method of classifying using the other feature amount.

(Third Embodiment) FIG. 31 shows the third embodiment of the present invention.
FIG. 6 is a diagram for explaining the processing of the image classification device in the embodiment of FIG.

Since the apparatus configuration of the image classification apparatus of the present invention is the same as that of the first embodiment, its explanation is omitted.

The processing performed in the image classifying apparatus in this embodiment is as follows.

After the processes of steps S201 to S207 of FIG. 2 are executed, the process of FIG. 31 is executed.

That is, in step S501, the feature amount indicating the character likeness of the effective connected component is extracted. The feature amount is extracted for all effective connected components in the image or for each effective connected component. The feature amount is data having one value in the entire image. The feature amount representing the character is the multi-valued image memory 1231, the binary image memory 12
32, label image memory 1233, fillet coordinate memory 1
234 is obtained.

In step S502, the processing device compares the extracted feature amount with a predetermined threshold value. The predetermined threshold value may be recorded in the ROM or may be externally designated by the user. Further, two or more feature amounts obtained by two or more different methods may be used for comparison.

In step S503, it is determined from the compared feature amount and the predetermined threshold value whether the feature amount satisfies the condition indicating the character-likeness. If YES, it is determined in step S504 that the input image is a document. To be done.

If NO in step S503, the input image is determined to be a non-document in step S506.

In step S505, the input image is recorded in the image database, but the keyword (attribute) "document image" is determined to be non-document for the image determined to be the document image at that time. The keyword "non-document image" is recorded together with the image.

As a result, when searching image data from the image database, keywords can be used to perform an efficient search.

(Fourth Embodiment) FIG. 32 shows a fourth embodiment of the present invention.
3 is a block diagram showing a system configuration of an image classification device in the embodiment of FIG.

Referring to the figure, the image classification device is roughly composed of an image input device 51 for inputting an image, which includes a scanner for a camera and the like, and an image by processing the input image as a business card image and other images. The image processing apparatus 52 that classifies (adds keywords) to, and the image database apparatus 53 that stores the images classified by the image processing apparatus 52. The respective devices are connected by a bus 15. The image processing device 52 includes a control device 521 including a CPU for processing and classifying an input image and controlling the operation of the entire system, a ROM 522 storing various programmable and constants, an image and the like. A RAM 523 that stores processing data and the like is included.

The ROM 522 includes an image processing unit 5221 used for image input processing, an image binarization processing unit 5222 used for binarizing an image, and binary image data among the binarized image data. A region extraction processing unit 5223 used for identifying a region, a labeling processing unit 5224 for performing labeling (processing for assigning a label number) to a connected component of data in binarized image data, and a region of image data A feature extraction processing unit 5225 for arranging regions for extracting the feature of the arrangement manner, an image classification processing unit 5226 for classifying images, and an image registration processing unit 5227 for registering image data in the image database device 53 are included.

The RAM 523 is a multi-valued image memory 5231 that stores the input image data as image data of 256 gradations, and a binary image memory that stores the image data represented by the binary values "0" and "1". 5232, a label image memory 5233 that stores the label number of each connected component of the data, a fillet coordinate memory 5234 that stores the fillet coordinates and area of each connected component, and the coordinates of the centroid of the effective connected component. Center of gravity memory 5235
A direction frequency histogram memory 5236 that stores the direction frequency between the center of gravity point and another center of gravity point, and a multivalued image memory 1231
Image memory 52 for storing the differential image of the image data of
37 and a density projection histogram memory 5 for storing a density histogram obtained by projecting the density of the differential image in the X and Y directions.
238, an effective connected component number memory 5239 for storing the effective connected component number, and a histogram differential value memory 5240 for storing the differential value of the histogram.

With this system configuration, the image input device 5
The various images input from 1 are automatically classified as business card images by the feature amount of the images being checked by the image processing device 52, and are stored in the image database device 53.

As a standard for defining the feature quantity in this embodiment,
The business card image uses the feature that the number of characters is small and the line spacing is wide.

33 to 35 are flow charts showing the processing of the control device 521 in this embodiment.

The flow chart shows a series of processes connected by the portion A or B in the figure. Steps S601 to S6
The processing in 07 is performed from steps S201 to S207 in FIG.
Since the processing is substantially the same as the processing described in step 1, the description thereof will not be repeated here.

At step S608, the controller 52
1 measures the number of effective connected components contained in the binary image memory 5232, and counts the number of effective connected component memories 5
Record at 239. Since the business card image has a small number of characters, it has a feature that the number of effective connected components is small. Therefore, the number of effective connected components is measured as one of the visual targets for determining the business card image.

In step S609, the centroid points of the respective effective connected components are determined and recorded in the centroid memory 5235.

The barycentric coordinates are obtained for the fillet coordinates for each valid connected component recorded in the fillet coordinate memory 5234.

If the upper left fillet coordinate of each effective connected component is (x_sp [i], y_sp [i]) and the lower right fillet coordinate is (x_ep [i], y_ep [i]),
The barycentric coordinate grav [i] is calculated by the equation (3).

Grav [i] = {(x_ep [i] −x_sp [i]) / 2, (y_ep [i] −y_sp [i]) / 2} (3) The centroid memory 5235 is shown in FIG. As shown, the X-coordinate and Y-coordinate of the center of gravity of the label are recorded in the order of the label numbers of the effective connected components.

The barycentric points for the connected components removed in step S607 are not recorded, so the label numbers are not necessarily continuous.

Further, since the memories for recording the X and Y coordinates of the center of gravity each have a memory capacity of 16 bits, a 32-bit memory area is secured for one label.

In step S610, the controller 52
1 checks other centroid points closest to each centroid point and records the direction in the direction frequency histogram memory 5236.

FIG. 37 is a flow chart showing a specific routine of the processing of the center of gravity performed in step S610 of FIG.

In step S801, the value of the variable "size" representing the size (pixel unit) of the search range is set to 1.

At step 802, the value of the variable "ws" representing the rectangular width of the search range is set by the equation ws = size × 2.

In step S803, it is determined whether or not there is another center of gravity on the side of the rectangle of one side ws centered on the center of gravity point of interest.

In step S804, there is no center of gravity point (N
If it is determined to be O), the variable “size” is incremented by 1 in step 809.

In step S810, it is determined whether the value of the variable "size" is smaller than the maximum threshold value MAX_SIZE. The value MAX_SIZE is set to 100, for example.

If YES in step S810, the processes from step S802 are repeated.

If YES at step S804, the direction from the center of gravity point at the center to the center of gravity on the side is calculated at step S805. At this time, if there are a plurality of center of gravity points on the rectangle, the directions are calculated for all the center of gravity points. The direction θ is calculated by the equation (4), assuming that the direction vector from the center to the center of gravity on the rectangle is (dx, dy).

Θ = tan ⁻¹ (dx / dy) (4) FIG. 38 shows a state in which another centroid point is found in the lower right of the centroid point of interest in the state of size = 3 and ws = 6. It is the figure shown. The shaded area in the figure indicates a rectangle with one side 6.

The direction vector from the center of gravity point to the center of gravity on the rectangle is represented by the difference between the coordinates of the respective centers of gravity, and in this example, (dx, dy) = (3,2), and θ = ta
n ⁻¹ (dx / dy) ≈0.98 [rad].

At step S806, the obtained direction is quantized and the direction frequency histogram memory 5236 is quantized.
Recorded in.

The range of quantization is as shown in FIG. 39, where θ is -π with reference to the horizontal direction (X-axis direction).
The area between / 2 and π / 2 (rad) is divided into four areas 0 to 3 each having an angle of π / 4. The determined direction is classified and recorded in any of these areas. For example, the direction vector θ≈ in FIG.
Since it is 0.98, it is classified into the region 3.

When the direction vector is not classified into the area of FIG. 39 (θ> π / 2, θ <−π / 2), the vector with the vector direction reversed is counted.

Therefore, in the present embodiment, the frequency histogram memory 5236 has four memory areas for counting the direction frequencies of the areas 0 to 3, respectively. Each memory area has 16 bits.

In step S807, the process proceeds to the next center of gravity point. In step S808, it is determined whether or not the directions of the recent centroid points have been checked for all the centroid points.
If it is ES, this routine is ended.

If NO in step S808, the processes from step S801 are repeated. N in step S810
If it is O, the process proceeds to step S807.

By the above processing, the cumulative frequency in the direction of the closest barycentric point for all barycentric points is recorded in the direction frequency histogram memory 5236.

In step S611 of FIG. 34, the control device 521 stores the multi-valued image memory 5231 in the multi-valued image memory 5231 in order to extract the characteristics of the interval and number of character-like regions of the input image from the information on the arrangement of the edge portions. Spatial differentiation processing is performed on the image data to create a differential image of 1 pixel 8 bits. The created differential image is stored in the differential image memory 5237. The method of creating the differential image is the same as that of the first embodiment using the Sobel operator, and therefore description thereof will not be repeated here.

In step S612, the controller 52
1 creates a density projection histogram which is density data projected on the X and Y axes based on the differential image stored in the differential image memory 5237. The density projection histogram is stored in the density projection histogram memory 5238.

To project an image on a certain axis means to repeatedly add the densities of pixels along a straight line in a direction perpendicular to the axis and calculate the total by repeating the parallel movement of the position of the straight line. Thus, a one-dimensional density arrangement (waveform) is obtained. For example, assuming a circle whose density is constant in all parts as shown in FIG.
The density projection histograms on the axes and the Y-axis are semi-elliptical histograms as shown in the figure.

As shown in FIG. 41A, the density projection histogram memory 5238 has an address configuration in which the X-axis and Y-axis coordinates are arranged in order, and the density projection value 16 bits for each coordinate is a unit. Has a memory area corresponding to the sum of the numbers of X and Y coordinates (in this case, 512 + 432 = 944).

When the number of character strings in the image is small (that is, when there are many blank portions), the density projection histogram in the direction of the character string has little change and the value becomes small.

At step S613, the controller 52
1 differentiates the density projection histogram and records it in the histogram differential value memory 5240.

This processing is carried out in order to emphasize the change in the density projection histogram to make it easier to understand. The specific contents of this processing will be described below.

The control device 521, with respect to the density projection histograms on the X and Y axes, recorded in the density projection histogram memory 5238, differentiates each value x_hist '.
[X] and y_hist ′ [y] are calculated by the equations (5) and (6), and the histogram differential value memory 524
Record at 0.

X_hist '[x] = (x_hist [x + 1] -x_hist [x-1]) / 2, 1≤x≤510 (5) y_hist' [y] = (y_hist [y + 1] -y_hist [y- 1]) / 2, 1 ≦ y ≦ 431 (6) where x_hist [i] and y_hist [i] are data (density projected values) in the density projected histogram memory 5238 at the coordinate i, respectively.

The processing in step S613 will be described with reference to FIG. As shown in FIG. 41 (b), the histogram differential value memory 5238 takes the addresses arranged in the order of the X coordinate / Y coordinate, and sets the differential value 16 bits of the density projection value for each coordinate as one unit. And the number of Y-coordinates (512 + 432 = 944 in this case). As shown in FIG. 41, the controller 54
1 is, for example, X = in the density projection histogram memory 5238.
0 and X = 2, the differential value is calculated using Equations (5) and (6), and X = of the histogram differential value memory 5240.
Record at address 1. This calculation is performed for all values on the X and Y axes. Since the differential value calculation requires that the coordinates exist before and after the pixel of interest, the differential value is not calculated for the first and last coordinates (here, coordinates 0 and 511 or 431). That is, as shown in FIG. 41B, X in the histogram differential value memory 5240
No data is recorded at the respective addresses of = 0, X = 511, Y = 0, and Y = 431.

Generally, the number of characters in a business card image is small, but the characters are well arranged in one direction and the space between lines is large. Therefore, the business card image has a smaller number of effective connected components than other images, the degree of concentration of the arrangement of the effective connected components in the specific direction is high (the degree of concentration of the frequency of the arrangement direction of the center of gravity in the specific direction), and the differential The average value of the density projection histogram tends to be small. Based on these tendencies, it is determined whether or not the input image is a business card image.

In step S614, step S6
At 08, it is determined whether the number of valid connected components stored is less than or equal to a threshold value. The threshold value is set to a value of 200, for example, and is stored in the ROM 522.

If YES in step S614, it is determined in step S615 whether the frequency of the direction from the center of gravity to the center of gravity closest to the center of gravity is greater than or equal to a threshold value. The threshold value is set such that the ratio of the mode value of the direction vector to the number of effective connected components is 50%.

If YES in step S615, the histogram differential value memory 524 in step S616.
The average value of the histogram recorded in 0 in the X-axis direction,
It is determined whether the larger or smaller average value in the Y-axis direction is less than or equal to the threshold value. The predetermined threshold is
For example, a value of 4.5 × 10 ³ is set for the larger average value, and a value of 1.5 × 10 ³ is set for the smaller average value and stored in the ROM 522.

Although each threshold value is stored in the ROM 522, the threshold value may be freely set according to the judgment of the user. In that case, the setting processing routine may be stored in the ROM 522.

If YES in step S616, it is determined in step S617 that the input image is a business card image.

If NO in any of steps S614, S615, S616, it is determined in step S619 that the input image is not a business card image.

In step S618, the input image is recorded in the image database device 53 together with the keyword (attribute) indicating whether it is a business card image.

As a result, at the time of processing when searching for each pixel,
You can use the attribute of the given business card. Further, by applying character recognition to an image determined to be a business card, an application such as address book creation is possible.

Next, a specific example of the differential value of the differential density projection histogram of the business card image will be described.

FIG. 44 is a graph showing an example of the differential value of the differential density projection histogram of the business card image, and FIG. 45 is a graph showing an example of the differential value of the differential density projection histogram of the document image other than the business card.

Referring to the drawing, the business card image has a larger average value of 2.26 × 10 ³ and a smaller average value of 0.86 × 10 ³ among the average values of the density projection histograms for the X axis and the Y axis.
For document images other than business cards, the larger one of the average values of the density projection histograms for the X axis and the Y axis is 4.3 × 10 ³ ,
The smaller one is 2.5 × 10 ³ .

As described above, it can be seen that the average value of the differential values of the differential density projection histogram is smaller in the business card image than in other document images.

Next, the actual classification result is shown in FIG. The X-axis of the graph of FIG. 42 is the number of effective connected components of the input image, and the Y-axis is the frequency ratio of the effective connected components of the input image in the direction of the nearest centroid.

As is clear from the figure, the business card image tends to have a small number of effective connected components and a high frequency ratio in the direction of the center of gravity.

The graph of FIG. 43 is only for the sample images included in the shaded area in the graph of FIG. 42 (the number of effective connected components is 200 or less and the frequency ratio in the direction of the center of gravity is 50% or more recently). Is a graph. The X axis of the graph is the smaller value of the average values of the differential values of the differential density projection histogram with respect to the X axis and Y axis of the input image, and the Y axis of the graph is the larger value. In this example, it is possible to classify the business card images by setting the threshold values of 1.5 × 10 ³ for the X axis and 4.5 × 10 ³ for the Y axis in FIG. Recognize.

As described above, according to this embodiment, the characteristics of the business card image can be utilized to accurately classify the business card image from the density projection histograms of the effective connected component image and the differential image. In addition, when it is determined that the image is a business card image, other applications such as address book creation and applications such as automatically proceeding to processing are possible.

In the present embodiment, three values, that is, the number of connected components effective for the determination of the business card image, the direction of other effective connected components that are close to the effective connected component, and the average of the differential values of the differential density projection histogram are used. Although the data is used, the determination can be made even if the determination is made by the data of 1 or 2 among them, although the accuracy is lowered.

(Fifth Embodiment) FIG. 46 shows the fifth embodiment of the present invention.
6 is a flowchart showing the processing of the control device of the image classification device in the embodiment of FIG.

Since the device configuration of this embodiment is the same as the device configuration of the image classification device of the fourth embodiment shown in FIG. 32, the description thereof will not be repeated here.

The processing performed by the image classification apparatus in this embodiment is as follows.

After the processing of steps S601 to S607 in FIG. 33 is performed, the routine shown in FIG. 46 is executed.

In step S608, the controller 52
1 regards the effective connected component in the image as a character-like area, calculates the number of effective connected components in the binary image memory 5232 as one of the feature amounts for classifying the business card image,
It is stored in the effective connected component number memory 5239. Further, in step S609, the control device 521 extracts the characteristic amount of the effective arrangement of the connected components in the binary image memory 5232. The feature quantity of the arrangement of the effective connected components is, for example, the degree of concentration of the arrangement of the effective connected components in a specific direction or the density projection of the differential image of the input image as described in the fourth embodiment of the present invention. A feature amount in which the characteristics of the business card image often appear, such as the average value of the differential value of the histogram, is used.

In step S610, in order to determine whether the input image is a business card image, the control device 521 first checks whether or not the number of valid connected components is less than or equal to a threshold value. Since the number of characters in the business card image is small, the number of effective connected components is also small. If the number of valid connected components is not less than or equal to the threshold value (NO in step S610), it is determined that the input image is not a business card image in step S614.

If the number of valid connected components is less than or equal to the threshold value (YES in step S610), the feature amount of the arrangement of valid connected components is compared with a predetermined threshold value in step S611. The predetermined threshold value may be stored in the ROM 522, or may be externally designated by the user. Further, the feature quantity to be compared may be a single feature or a combination of a plurality of feature quantities. As the predetermined threshold value, a single threshold value or a combination of a plurality of threshold values is used corresponding to the feature amount to be compared, and the threshold value condition that the document image should satisfy is set.
Those processing procedures are stored in the ROM 522.

In step S612, the controller 52
As a result of the comparison between the feature amount and a predetermined threshold value 1, it is determined whether or not the input image satisfies the condition for being a business card image. The conditions are met (YES in S612)
In this case, the input image is determined to be a business card image,
Otherwise (NO in step S612), it is determined that the input image is not a business card image.

In step S615 after the determination, the control device 521 registers the input image in the image database device 53, and the keyword (attribute) "business card image" is added to the image determined to be the business card image at that time. It is registered in the image database device 53 together with the input image data. As a result, the attribute of a business card that has been given can be used during the process of searching each image.

As described above, according to the present invention, the number of characters can be reduced by checking the number of character-like areas and the arrangement of the areas and determining whether or not the condition of the business card image is satisfied in the entire image. The feature of the business card image that the line spacing is wide open can be extracted and classified accurately.

Further, when classified as a business card image, it is possible to automatically create an address book by applying character recognition.

Although the image data input in the embodiment of the present invention is multi-valued image data, binary image data may be input.

[0193]

The image classification device of the present invention can classify images into business card images and other images.

In addition to the above effects, it is possible to perform image classification with higher determination accuracy.

In addition to the above effects, a pixel satisfying a predetermined condition among a plurality of adjacent pixels representing an object is identified as a connected component, so that the determination accuracy can be further improved.

[Brief description of drawings]

FIG. 1 is a block diagram showing a system configuration of an image classification device according to a first embodiment of the present invention.

FIG. 2 is a flowchart showing a processing routine of a control device 121 of FIG.

FIG. 3 is a flowchart following the flowchart of FIG.

FIG. 4 is a diagram showing an example of a document image.

FIG. 5 is a diagram illustrating an example of a landscape image that is a non-document image.

FIG. 6 is a diagram for explaining a coordinate system of image data stored in a multi-valued image memory 1231.

FIG. 7 is a diagram for explaining an address configuration of a multi-valued image memory 1231.

FIG. 8 is a diagram showing an example of binary image data and explaining connected components.

9 is a diagram showing the results of labeling the connected components of FIG.

FIG. 10 is a diagram showing an address configuration of a label image memory.

FIG. 11 is a diagram for explaining a method for setting fillet coordinates.

FIG. 12 is a diagram showing an address configuration of a fillet coordinate memory 1234.

FIG. 13 is a diagram showing an example of connected components.

FIG. 14 is a diagram showing a state in which the connected component of FIG. 13 is recorded in a fillet coordinate memory.

15 is a diagram showing a connected component after unnecessary connected components are removed from FIG. 13. FIG.

FIG. 16 is a diagram for describing data for performing spatial differentiation processing on multivalued image data in a multivalued image memory.

FIG. 17 is a first diagram illustrating a coefficient used for differentiation.

FIG. 18 is a second diagram for explaining a coefficient used for differentiation.

19 is a diagram showing a state in which one connected component in the document image of FIG. 4 is recorded in a binary image memory 1232.

20 is a diagram showing a state in which the same connected component as the connected component shown in FIG. 19 is recorded in a differential binary image memory 1234. FIG.

FIG. 21 is a diagram for explaining the same connected component as in FIG. 19 in a state of being recorded in the AND image memory 1237.

22 is a diagram showing a state in which one connected component in the landscape image of FIG. 5 is recorded in a binary image memory 1232.

23 is a diagram showing a state in which the same connected component as the connected component shown in FIG. 21 is recorded in a differential binary image memory 1234. FIG.

FIG. 24 is a diagram for explaining the same connected component as that in FIG. 23 in a state of being recorded in the AND image memory 1237.

FIG. 25 is a diagram for explaining an implementation result (including effects) in the first example of the present invention.

FIG. 26 is a block diagram showing a device configuration of an image classification device according to a second embodiment of the present invention.

FIG. 27 is a flowchart showing a process performed by the control device of the image classification device in the second embodiment.

FIG. 28 is a diagram showing an example of binary image data in a document image.

FIG. 29 is a diagram showing an example of binary image data in a landscape image (non-document image).

FIG. 30 is a diagram for explaining a bin ratio of a document image and a non-document image.

FIG. 31 is a diagram for explaining the process of the image classification device in the third embodiment of the present invention.

FIG. 32 is a block diagram showing a system configuration of an image classification device according to a third embodiment of the present invention.

FIG. 33 is a flowchart showing a process performed by the control device 521 of the image classification device according to the fourth embodiment of the present invention.

FIG. 34 is a flowchart following FIG. 33.

FIG. 35 is a flowchart following FIG. 34.

FIG. 36 is a diagram showing an address configuration of a centroid memory 5235.

37 is a flowchart showing a specific routine of processing of a center of gravity performed in step S610 of FIG. 34.

38 is a diagram showing a state in which another centroid point is found in the lower right of the centroid point of interest in the state of size = 3, ws = 6 in the process of FIG. 37. FIG.

FIG. 39 is a diagram for explaining a region for quantizing the obtained direction.

FIG. 40 is a diagram for explaining an example of density projection of image data on the X axis and the Y axis.

41 is a diagram for explaining the density projection histogram memory 5238 and the histogram differential value memory 5240. FIG.

FIG. 42 is a diagram for explaining an implementation result (including effects) in the fourth example of the present invention.

43 is a diagram for explaining the average value of the differential values of the differential density projection histogram of the image data included in the shaded area of FIG. 42.

FIG. 44 is a diagram for explaining a density projection histogram in the X-axis direction and the Y-axis direction for a business card image.

FIG. 45 is an X-axis direction and Y of a document image other than a business card image.
It is a figure for demonstrating the density projection histogram of an axial direction.

FIG. 46 is a flowchart showing the processing of the control device in the fifth embodiment of the present invention.

[Explanation of symbols]

11,51 image input device 12,52 Image processing device 13,53 Image database device 121,521 control device 122,522 ROM 123,523 RAM 1231, 5231 Multi-valued image memory 1232, 5232 binary image memory 1233, 5233 Label image memory 1234, 5234 Fillet coordinate memory 1235, 5237 differential image memory 1236 differential binary image memory 1237 AND image memory 5235 Centroid memory 5236 Direction frequency histogram memory 5238 Density projection histogram memory 5239 Effective connected component number memory 5240 Histogram differential value memory

─────────────────────────────────────────────────── ─── Continuation of front page (58) Fields surveyed (Int.Cl. ⁷ , DB name) G06T ⁷ /00-7/60 G06T 1/00 G06F 17/30 170 G06K 9/20 H04N 1/40

Claims (6)

(57) [Claims] 1. An identification unit for identifying a connected component composed of a plurality of adjacent pixels representing an object from input image data, a measuring unit for measuring the number of the connected components as a feature amount, and the measurement. An image classification device, comprising: a determination unit that determines whether or not the input image data is a business card image based on the obtained feature amount. 2. A discriminating means for discriminating a connected component consisting of a plurality of adjacent pixels representing an object from the input image data, and dividing the discriminated connected component for every two adjacent connected components. A direction extracting unit that extracts a line-up direction for each of the two connected components that have been divided; An image classification device, comprising: a determination unit that determines whether the input image data is a business card image based on the feature amount. 3. A differential image data creating means for creating differential image data from input image data, a density projection histogram creating means for creating a density projection histogram based on the differential image data, and the created density. An image classification device comprising: a calculation unit that calculates a feature amount from a projection histogram; and a determination unit that determines whether or not the input image data is a business card image based on the calculated feature amount. 4. The two nearest neighbors of the identified connected component
A direction extraction unit that divides each connected component and extracts the arrangement direction of each of the two connected components, and a measurement that measures the degree of concentration of the extracted arrangement direction in a specific range as a feature amount. Further comprising means, wherein the feature amount includes a degree of concentration in the direction of the specific range,
The image classification device according to claim 1. 5. A differential image data creating means for creating differential image data from the input image data, a density projection histogram creating means for creating a density projection histogram based on the differential image data, and the created The image classification apparatus according to claim 1, further comprising: a calculating unit that calculates data based on a differential value of the density projection histogram, wherein the feature amount includes data based on a differential value of the density projection histogram. 6. The method according to claim 4, wherein the identification means identifies, from the input image data, a pixel satisfying a predetermined condition among a plurality of adjacent images representing an object as the connected component. The image classification device according to any one of the above.