JP4116377B2 - Image processing method and image processing apparatus - Google Patents

Description

[0001]
BACKGROUND OF THE INVENTION
The present invention relates to an image processing method and an image processing apparatus for identifying, in particular, a photograph area from a document image in which characters, photographs, diagrams, tables, and the like are mixed.
[0002]
[Prior art]
In recent years, document images are increasingly handled with multi-valued images such as color and gray scale, so area identification, title character string extraction, which has been performed mainly for binary document images, In processes such as table processing and OCR, there is a demand to execute on multi-valued images. Since various methods already exist for binary document images, it is considered that existing processing for binary images can be applied if the multi-value document image is binarized well.
[0003]
FIG. 6 shows a document image. On the document image 20, areas having different attributes such as a character area 21, a table area 22, a figure area (not shown), and a photograph area 23 are mixed. Among these, in the character area 21, the table area 22, and the figure area, the number of colors used for describing them is often small. It is rare that one character itself is described in multiple colors, or one cell of a table is multicolored. Therefore, for these regions, it is possible to partially find an appropriate threshold value, binarize, and use the existing processing for the binary image.
[0004]
However, in the photo area 23, depending on the content of the photo, the types of pixel colors are much larger. If this photo area 23 is binarized by the same threshold processing as other areas, not only will the contents of the photo be indistinguishable, but other errors may occur while processing the binarized image. It was sometimes classified into the area of.
[0005]
As a method of area classification for such a multi-valued image, a background color is extracted, partial areas composed of other colors are sequentially classified into ruled lines and character areas, and finally they are not classified as either. There is a method in which the remaining area is set as the other area, and this other area is handled as a figure or photo area (see, for example, Patent Document 1).
[0006]
Patent Document 2 relates to a technique for specifying a background color of an image to be described later.
[0007]
[Patent Document 1]
JP 2001-297303 A [Patent Document 2]
JP 2001-222717 A
[Problems to be solved by the invention]
However, the conventional technique described above is not suitable for the purpose of extracting a photograph because if there is an array of rectangles in a shape similar to a character string in the photograph region, it is classified as a character region.
[0009]
In order to solve the above-described problems caused by the prior art, the present invention effectively utilizes the shape information of the components constituting the region in addition to the pixel color information for the multi-valued image, thereby reducing the number of identification errors. An object of the present invention is to provide an image processing method and an image processing apparatus capable of identifying a region.
[0010]
[Means for Solving the Problems]
In order to solve the above-described problems and achieve the object, an image processing method according to the present invention is an image processing method having a function of identifying a photographic region from a document image input as a multi-valued image. A background color specifying step for specifying a background color of an image, a partial region extracting step for extracting a partial region in which pixels of a color different from the background color are connected, a series of pixels of the same color adjacent to the extracted partial region, etc. Color matching connected component extraction step for extracting as a color connected component, circumscribed rectangle of the equal color connected component in the partial area, a connected component feature extracting step for obtaining a plurality of circumscribed rectangles overlapping, and overlapping of the circumscribed rectangles And a photographic region determination step for identifying whether or not the partial region is a photographic region.
[0011]
In addition, the method includes recursively identifying a photographic area with respect to a partial area identified as a photographic area in the photographic area determination step, by regarding the partial area as an input image and executing the series of processes a predetermined number of times. It is characterized by that.
[0012]
The image processing apparatus according to the present invention is an image processing apparatus having a function of identifying a photographic region from a document image input as a multi-valued image, and a background color specifying means for specifying a background color of the document image And a partial area extracting means for extracting a partial area in which pixels of a color different from the background color are connected, and a uniform color connected component extraction for extracting a series of pixels of the same color adjacent to the extracted partial area as a uniform color connected component Means for obtaining a circumscribed rectangle of the same color connected component in the partial area and obtaining a plurality of circumscribed rectangles overlapping, and whether the partial area is a photographic area based on the overlapping of the circumscribed rectangles Photographic area determination means for identifying whether or not.
[0017]
DETAILED DESCRIPTION OF THE INVENTION
Exemplary embodiments of an image processing method and an image processing apparatus according to the present invention are explained in detail below with reference to the accompanying drawings. FIG. 1 is a block diagram showing the configuration of an embodiment of an image processing apparatus of the present invention.
[0018]
The image processing apparatus includes an image input unit 1 that inputs a multivalued document image using an image input device such as a color scanner, a background color extraction unit 2 that specifies a background color of the input multivalued document image, and a background A partial region extraction unit 3 that extracts a region in which pixels different from color are connected, a color matching component extraction unit 4 that extracts a series of adjacent pixels of the same color in the partial region, and colors and shapes of the color matching connected components A connected component feature extracting unit 5 for extracting features, a photo region determining unit 6 for determining whether a partial region image is a photo region from the obtained connected component features, a memory 7 for storing the extracted data, and processing It consists of the control part 8 which controls the flow of this.
[0019]
(Embodiment 1)
FIG. 2 is a flowchart showing a flow of image processing according to the first embodiment. First, an image is input (step S1), and a background color is specified for the input document image (step S2). The background color can be specified using a method disclosed in Patent Document 2, for example.
[0020]
Next, a partial region in which pixels other than the background color are connected is extracted (step S3). Extraction of connected pixels is a widely known process, and various methods can be used. Next, a uniform color connected component is extracted from the partial region (step S4). The uniform color connected component is obtained by repeating the operation of grouping the adjacent pixel colors as the same color if they are within a certain distance in the color space. Thus, the pixels in the partial area belong to any connected component.
[0021]
Then, the feature of the color matching connected component in this partial region is extracted (step S5). As a feature, the number of color types of the equal color connected components constituting the partial area is obtained. The number of color types is compared with a predetermined threshold value, and if it is greater than the threshold value, the partial area is determined as a photographic area (step S6). The threshold value may be obtained experimentally.
[0022]
In general, there are often only one type of pixel color that constitutes a character, or several types even if it is an outline or a shaded character. On the other hand, a photograph is composed of far more kinds of pixel colors. Also, a table in which a cell has a color is more than the color of a pixel constituting a character, but is often a solid color combination, and the number of pixel colors is not so large as in a photographic area. As described above, if the number of colors of the equal color connected components constituting the partial area is used, a photograph area can be easily extracted from a multi-value image with high accuracy.
[0023]
(Embodiment 2)
In the second embodiment, in the feature of the color matching connected component in the partial area described in the first embodiment (step S5), the number of the color matching connected components is also obtained and used for the determination of the photographic area. In general, in a photographic region, color matching components are often finely divided compared to characters and character backgrounds.
[0024]
For example, in an aerial photograph of an urban area, if a color matching component is extracted for each building or car, a very large number of color matching components exist. If the color types of these color matching components are similar, the number of color matching components becomes enormous, although the number of color types is small, it can be distinguished from other regions. In this way, by using together the number of color matching components constituting the partial area, it is possible to easily extract even a photographic area having a small number of color types.
[0025]
(Embodiment 3)
In the third embodiment, in the feature of the color matching connected component in the partial area described in the second embodiment (step S5), the feature regarding the overlapping of the circumscribed rectangles of the color matching connected component is also obtained and used for the determination of the photographic area. FIG. 3 is a diagram for explaining the circumscribed rectangle distribution of the equal color connected components.
[0026]
In general, in the photographic region shown in FIG. 3C, the shape of the color matching connected component is often not rectangular compared to characters and the background of characters. Therefore, as shown in FIG. 3D, the circumscribed rectangles of adjacent color matching connected components greatly overlap each other. On the other hand, in the character region shown in FIG. 3A, the color matching components constituting the character may overlap to some extent, but adjacent characters hardly overlap each other. Thus, there is little overlap of circumscribed rectangles.
[0027]
FIG. 4 is a diagram for explaining the overlapping state of circumscribed rectangles. However, since it overlaps with the same color connected components corresponding to the background such as characters, it is necessary to eliminate the overlap with the background. As shown in FIG. 4 (a), the overlap of the background color matching components with the circumscribed rectangle is a complete inclusion relationship. Therefore, as shown in FIG. 4B, the overlapping of circumscribed rectangles of the equal color connected components that are not completely inclusive is examined, and the ratio of the total area of the overlapping portions to the area of the partial regions exceeds a certain threshold value. It can be distinguished from other areas as a photograph.
[0028]
In this way, even in a photographic area where the number of types of colors is small and the number of equal color connected components is similar to that of a character area by using the overlapping of circumscribed rectangles of the same color connected components constituting the partial area. Can be easily extracted.
[0029]
(Embodiment 4)
In the fourth embodiment, in the feature of the color matching connected component in the partial region described in the third embodiment (step S5), the feature relating to the distribution of the size of the color matching connected component is also obtained and used for the determination of the photographic region. In general, photographic areas often have different sizes of equal color connected components as compared to areas containing characters. For example, a human face photograph is divided into face outlines, eyes, mouths, hair, clothes, backgrounds, etc., but their sizes are not distributed within a certain range like character strings.
[0030]
Accordingly, the variance value of the area of the circumscribed rectangle of the equal color connected component is obtained, and if this value exceeds a certain threshold value, it can be distinguished from other regions as a photograph. Instead of the area, dispersion of the length of the long side or the short side of the circumscribed rectangle can also be used. In this way, by using together the distribution of the sizes of the equal color connected components constituting the partial area, it is possible to easily extract even a photographic area where the number of color types is small and the circumscribed rectangles are not overlapped. Can do.
[0031]
(Embodiment 5)
FIG. 5 is a flowchart showing the flow of processing in the fifth embodiment. After the photographic area is determined in step S6 by the processing of the first to fourth embodiments described above, the determiner (step S7) regards it as a newly input image, returns to step S1, and recursively sets the photographic area. Let them be identified. The number of recursive processes in step S7 may be determined in advance, for example, twice. This makes it possible to correctly extract only the photographic area even when the photographic area is included in another area different from the background on the document image.
[0033]
As described above, according to the image processing method according to the present embodiment, not only the pixel color information but also the shape information of the components constituting the region is effectively used for the multi-valued image, so that the identification is performed. Photo areas with few errors can be identified.
[0034]
The image processing method described in the present embodiment can be realized by executing a program prepared in advance on a computer such as a personal computer or a workstation. This program is recorded on a computer-readable recording medium such as a hard disk, a flexible disk, a CD-ROM, an MO, and a DVD, and is executed by being read from the recording medium by the computer. The program may be a transmission medium that can be distributed via a network such as the Internet.
[0035]
【The invention's effect】
As described above, according to the present invention, it is possible to identify a photographic region with few identification errors by effectively using shape information of components constituting the region in addition to pixel color information for a multi-valued image. There is an effect that it is possible. The shape information includes the number of types of color of the same color connected component of the partial area, the number of the same color connected component of the partial area, the overlapping state of the circumscribed rectangles of the same color connected component of the partial area, and the same color connected of the partial area. This is done using the component size distribution.
[0036]
Also, by recursively identifying the photo area or changing the number of times the photo area is extracted recursively according to the image, even if the photo area is included in other areas, the photo area is correctly The effect is that only can be extracted.
[Brief description of the drawings]
FIG. 1 is a block diagram showing a configuration of an embodiment of an image processing apparatus of the present invention.
FIG. 2 is a flowchart showing a flow of image processing according to the first embodiment.
FIG. 3 is a diagram for explaining a circumscribed rectangle distribution of equal color connected components;
FIG. 4 is a diagram for explaining an overlapping state of circumscribed rectangles.
FIG. 5 is a flowchart showing a flow of processing in the fifth embodiment.
FIG. 6 is a diagram illustrating a document image.

Claims (3)

An image processing method having a function of identifying a photographic region from a document image input as a multi-valued image,
A background color specifying step for specifying a background color of the document image;
A partial region extraction step of extracting a partial region in which pixels of a color different from the background color are connected;
A color matching connected component extraction step for extracting a series of pixels of the same color adjacent to the extracted partial region as a color matching connected component;
Obtaining a circumscribed rectangle of equal-color connected components in the partial region, and obtaining a connected component feature extracting step for obtaining an overlap of the circumscribed rectangles ;
A photograph region determining step of identifying whether the partial area based on the overlap of the circumscribed rectangle is a photograph region,
An image processing method comprising: Recursively identifying a photographic area with respect to a partial area identified as a photographic area by the photographic area determination step, including the step of considering the partial area as an input image and executing the series of processes a predetermined number of times. The image processing method according to claim 1, wherein: An image processing apparatus having a function of identifying a photographic region from a document image input as a multivalued image,
Background color specifying means for specifying the background color of the document image;
Partial area extraction means for extracting a partial area in which pixels of a color different from the background color are connected;
Color matching connected component extraction means for extracting a series of pixels of the same color adjacent to the extracted partial area as a color matching connected component;
A connected component feature extracting means for obtaining a circumscribed rectangle of the equal color connected component in the partial region and obtaining an overlap of the circumscribed rectangle;
Photographic area determination means for identifying whether the partial area is a photographic area based on the overlap of the circumscribed rectangles;
An image processing apparatus comprising:

Images