JP2000067248A - Picture matching method/device - Google Patents

Abstract

PROBLEM TO BE SOLVED: To make it possible to perform retrieval under a user's ambiguous memory by performing processing for integrating each area so that each area included in pictures can be in one-to-one corresponding between the pictures. SOLUTION: A picture which a user prepares for inquiry is inputted by an input means. An area picture is similarly inputted from within a picture database. Next, an area division picture normalizing means receives the area picture from the input means, and normalizes the picture up to the same picture size by contracting or magnifying one or both of the pictures (a step 1). An area integration means receives the normalized pictures from the area division picture normalizing means and integrates areas so that the areas divided by both pictures can be clearly coordinated (a step 2). An area compatibility degree leading means receives the picture after integrates from the area integration means and leads the number of areas and the degree of compatibility of a shape every area.

Description

DETAILED DESCRIPTION OF THE INVENTION

[0001]

BACKGROUND OF THE INVENTION 1. Field of the Invention The present invention relates to an image matching apparatus,
In particular, the present invention relates to an image matching method and apparatus used for searching in an image database system.

[0002]

2. Description of the Related Art In the field of image retrieval, there is a growing need for retrieval based on features unique to images. Among search requests based on the content of an image, there are requests related to information search based on the spatial characteristics of an object in the image.

When performing such a search, it is necessary to match the region image used for the inquiry with the region image in the database, and the similarity is determined based on the position information of the object and the shape information of the region of the object. are doing. Note that the region image referred to here is an image that is region-divided by a matrix based on feature amounts such as patterns, colors, areas, and contrasts.

[0004]

However, in many cases, the area image used for the inquiry and the area image in the database are created from different viewpoints. Over-partitioning of the regions occurs, which causes difficulties in actually matching. In addition, in order to perform matching in consideration of both the color information and the area information, it is necessary to take into consideration ambiguity of specification of both, lack of storage, and the like, and matching is difficult.

An object of the present invention is to solve such a problem and to realize an image matching method and apparatus capable of retrieving color information and shape information as clues even under ambiguous memory of a user.

[0006]

An image matching method according to the present invention is an image matching method for measuring a degree of matching between images obtained by dividing an area, wherein each area included in the image is interposed between the images. A process for integrating the respective regions so as to be associated one-to-one is performed, and a degree of matching between the images is measured for at least the integrated region using at least shape information.

The image matching method of the present invention is an image matching method for measuring the degree of matching between divided images, wherein each region included in the image has a one-to-one correspondence between the images. A process of integrating the respective regions so that the images can be attached is performed, and the degree of matching between the images is measured using at least both the shape information and the color information before the integration with respect to the integrated region.

The image matching apparatus according to the present invention is an image matching apparatus for measuring a degree of matching between divided images, wherein each area included in the image has a one-to-one correspondence between the images. A region integration unit that performs a process of integrating the respective regions so that the regions can be attached, and a region matching degree deriving unit that measures the degree of matching between the images using shape information for the integrated region. It is characterized by being performed.

An image matching apparatus according to the present invention is an image matching apparatus for measuring a degree of matching between divided images, wherein each area included in the image has a one-to-one correspondence between the images. Area integration means for performing processing for integrating the respective areas so as to be attached, and image matching for measuring the degree of matching between the images using at least both the shape information and the color information before integration for the integrated area And a degree deriving means.

[0010]

SUMMARY OF THE INVENTION The present invention is an image matching method and apparatus capable of matching fragmentary storage of image information with respect to colors and shapes and ambiguous queries.

[0011]

Embodiments of the present invention will be described with reference to the drawings.

FIG. 1 is a functional block diagram showing an embodiment of the present invention. FIG. 5 is a diagram showing an example of a region image, and FIG. 6 is a diagram for simply showing a process of processing.

In FIG. 1, an input means 1 inputs an image created by a user for an inquiry. As an example of the form of the region image, an image representing a matrix indicating the region and characteristic amounts (color, contrast) of the region as shown in FIG. 5 can be considered. Similarly, an area image is input from the image database. Next, the area division image normalizing means 2 receives the area image from the input means 1 and normalizes to one and the same image size by reducing or enlarging one or both images. (Step 1 in FIG. 6) The area integrating means 3 receives the normalized image from the area divided image normalizing means 2 and integrates the areas so that the areas divided by both images can be clearly corresponded. (Step 2 in FIG. 6)

The area matching degree deriving means 4 receives the integrated image from the area integrating means 3 and derives the number of areas and the degree of matching of the shape of each area. As an example, the number of pixels having the same coordinate area number is derived.

The output means 5 receives the matching result from the area matching degree deriving means 4 and outputs it.

Next, an embodiment of the present invention will be described with reference to the drawings.

FIG. 2 is a functional block diagram for implementing the present invention.

In FIG. 2, the user inputs an image created for inquiry by the input means 21. As an example of the form of the region image, an image showing a matrix indicating the region and a feature amount of the region as shown in FIG. 5 can be considered. Similarly, an area image is input from the image database. Next, the area dividing image normalizing means 22 is
More area images are received and one or both images are reduced or enlarged to normalize to the same image size. (Step 1 in FIG. 6) The area integrating means 23 receives the images normalized by the area divided image normalizing means 22 and integrates the areas so that the areas divided by both images can be clearly corresponded. (Step 2 in FIG. 6)

The area matching degree deriving means 24 includes the area integrating means 2
The integrated image is received from 3 and the number of regions and the degree of matching of the shape of each region are derived. As an example, the number of pixels having the same coordinate area number is derived.

The color information matching degree deriving means 25 includes an input means 21
Further, the color information of the area is received, and information on integration is received from the area integration means 23 to derive the degree of consistency of the color information.

The inter-image matching degree deriving means 26 receives the matching result from the area matching degree deriving means 24 and the color information matching degree deriving means 25, and derives the matching degree of the entire image by considering both.

As an example, the whole degree of consistency is determined by a linear combination of the degree of consistency based on the shape and the degree of consistency based on the color information, and the newly derived degree of consistency is set to a value previously given based on the number of integrated regions. There is a method of converting.

The output means 27 is an image matching degree deriving means 26
Receives and outputs more matching results.

Next, embodiments related to the present invention will be described with reference to the drawings.

FIG. 3 is a functional block diagram showing an embodiment related to the present invention. FIG. 7 is a flowchart showing the flow of the processing of this embodiment. FIG. 8 is a diagram showing the correspondence in this embodiment.

In FIG. 3, the input means 31 reads two area images. (Step 71 in FIG. 7) At this time, two images are assumed to be A and B. The pixel unit correspondence deriving means 32 derives a correspondence for each pixel of the two images. The values of the area of (m, n) (x coordinate is m, y coordinate is n) of the two area images are Amn and Bmn, respectively. A in image A
The region b in the image B is defined as a region b in the image B for all the pixels belonging to the region a in the image A (for all (m, n) where Amn = a).
Indicated as b. In this case, b is a reference link of a. Similarly, the area corresponding to the area b in the image B is described as a ← b. In this case, a is a reference link of b. The pixel unit correspondence deriving means 32 obtains corresponding areas for all areas in the image A, and obtains corresponding areas for all areas in the image B. (Steps 72 and 73 in FIG. 7) The pixel unit correspondence storage unit 33 stores this result. FIG. 8 shows the correspondence of the correspondence in the pixel unit correspondence storage unit 33.

The one-to-one correspondence area deriving means 4 finds a set of areas (a, b) where a → b and a ← b, and assigns a new area number to them. (Step 74 in FIG. 7) The area integrating means 35 performs area integration on the area to which a new area number has not been assigned by the one-to-one corresponding area deriving means 34.
Of the remaining areas, if the corresponding area has already been assigned a new area number, the process of setting the area number as a new area number is repeated until all the areas A and B are numbered. (Step 75 in FIG. 7) The output means 36 outputs the result according to the newly assigned area number. (Step 76 in FIG. 7)

Next, embodiments related to the present invention will be described with reference to the drawings.

FIG. 4 is a functional block diagram showing an embodiment related to the present invention. FIG. 9 is a diagram showing a method of calculating the degree of color matching in this embodiment.

In FIG. 4, an input means 41 inputs an area image after area integration (corresponding to each area on a one-to-one basis), an image before area integration, and color information on the area.

The area unit information extracting means 42 extracts the number of pixels and color information of each area before integration, which are variables necessary for deriving the degree of color matching for each area after integration.

The inter-color distance score deriving means 44 calculates and returns the similarity between two colors when they are input.
As an example, the hue and lightness components of two colors are represented by H1 and V, respectively.
1, H2, V2, and DIST is the distance between two vectors

(Equation 1) (Where i and j are appropriate weighting factors, const is a constant for keeping the score positive), and a method of changing the value of i depending on the magnitude of the saturation.

The area unit color matching means 43 receives the number of pixels and the color information of the area before integration from the area unit information extraction means 42, and sends the color information to the color distance score deriving means 44 to obtain the color distance. The degree of color matching for each area is derived by multiplying the score by the number of pixels.

The matching sum deriving means 45 receives the degree of matching for each area from the area unit color matching means 43 and adds them to derive the color matching degree of the entire image.

The output means 45 receives and outputs the result from the matching sum deriving means 45.

[0036]

As described above, according to the present invention,
Since the region image used for the inquiry and the region image in the database are created from different viewpoints, there is an effect that it is possible to solve a difficulty in matching due to a difference in memory of a user, forgetting of a region, excessive division of a region, and the like. is there. This makes it possible to make the matching at the time of retrieval ambiguous. In other words, the features that the user has forgotten at the time of the inquiry are collected at the integration stage even if they are confirmed in the database, so that there is an advantage that the feature is searched with a high score. Conversely, even if a feature that is not in the database is included in the user's input, the search is performed after integration, so that there is an effect that the search is not hindered. Further, in the present invention, since both the color information and the shape information are used efficiently, there is an effect that one of the information becomes ambiguous at the time of searching, and it is possible to search even if the information is forgotten in an extreme case. .

[Brief description of the drawings]

FIG. 1 is a functional block diagram showing an embodiment of the present invention.

FIG. 2 is a functional block diagram showing an embodiment of the present invention.

FIG. 3 is a functional block diagram showing an embodiment related to the present invention.

FIG. 4 is a functional block diagram showing an embodiment related to the present invention.

FIG. 5 is a diagram showing an example of a region image according to the present invention.

FIG. 6 is a diagram showing a process of processing according to the present invention.

FIG. 7 is a flowchart showing a processing flow of an embodiment related to the present invention.

FIG. 8 is a diagram showing a correspondence relationship in the embodiment related to the present invention.

FIG. 9 is a diagram showing a method of calculating the degree of color matching in an embodiment related to the present invention.

[Explanation of symbols]

 REFERENCE SIGNS LIST 1 input means 2 area division image normalization means 3 area integration means 4 area consistency derivation means 5 output means 21 input means 22 area division image normalization means 23 area integration means 24 area consistency degree derivation means 25 color information consistency degree derivation means 26 Derivation means between images 27 Output means 31 Input means 32 Pixel correspondence correspondence derivation means 33 Area correspondence storage means 34 One-to-one correspondence area derivation means 35 Area integration means 36 Output means 41 Input means 42 Area unit information extraction means 43 Area unit color matching means 44 Inter-color distance score deriving means 45 Matching sum deriving means 46 Output means

Claims (4)

[Claims] 1. An image matching method for measuring the degree of matching between images obtained by dividing an area, wherein the areas are integrated so that each area included in the image is associated one-to-one with the images. An image matching method for performing a process of measuring the matching degree between the images using at least the shape information for the integrated area. 2. An image matching method for measuring the degree of matching between images obtained by dividing an area, wherein the areas are integrated such that each area included in the image is associated one-to-one with the images. An image matching method, wherein the matching degree between the images is measured using at least both the shape information and the color information before the integration for the area after the integration. 3. An image matching apparatus for measuring a degree of matching between images obtained by dividing an area, wherein the areas are integrated so that each area included in the image is associated one-to-one with the images. An area integration means for performing a process of performing the processing, and an area matching degree deriving means for measuring the degree of matching between the images using shape information for the integrated area. Matching device. 4. An image matching apparatus for measuring the degree of matching between divided images, wherein the respective regions are integrated such that the respective regions included in the images are associated with each other on a one-to-one basis. Region integration means for performing a process of performing, and at least image-to-image matching degree deriving means for measuring the matching degree between the images using at least both the shape information and the color information before integration for the integrated area. An image matching device characterized by comprising: