JP3478627B2 - Image retrieval apparatus and method - 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 retrieval apparatus and method for retrieving similar images from an image database on the basis of input example images.

[0002]

2. Description of the Related Art Conventionally, when a necessary image is searched from an image database, a searcher's memory, for example, an image shown in FIG. , If there is a lump having a color close to that shown by the symbol A, the searcher sets the color, shape, and place of the lump as a closed figure as shown in FIG. Draw. And for every image in the image database,
There is a method of comparing the position and color of the graphic specified in this example image to calculate the similarity and displaying the images in descending order of similarity.

[0003]

However, in the above-described conventional search method, for example, even if the image stored in the searcher's memory has a clear edge as shown in FIG. When there is a gradation, it is difficult to determine how much of the image is the uniform color area due to the gradation, so there is a problem that it is difficult to specify the uniform color area as an example image.

The present invention has been made in view of the above problems, and an object of the present invention is to provide an image retrieving apparatus capable of retrieving a desired image only by inputting an edge position of the image and its peripheral color, and an image retrieving apparatus therefor. Is to provide a method.

[0005]

[Means for Solving the Problems] and

In order to achieve the above object, the present invention is an image retrieval apparatus for retrieving a desired image from a plurality of image data stored in a database, and means for inputting an edge line of an exemplary image of an image to be retrieved. A means for designating a color for the peripheral area of the edge line; a means for calculating the similarity between the example image and the plurality of image data from the colors of the edge line and the peripheral area; An image corresponding to the maximum image data is displayed as the searched image.

Further, another invention further comprises a calculating means for calculating a distance value between the exemplary image and the plurality of image data, and the distance value is used as the similarity.

The above-mentioned structure functions to search for a desired image quickly and surely.

[0008]

DETAILED DESCRIPTION OF THE PREFERRED EMBODIMENTS Preferred embodiments of the present invention will be described in detail below with reference to the accompanying drawings.

FIG. 5 is a block diagram showing the configuration of the image database system according to the embodiment of the present invention. In the figure, reference numeral 1 is the main body of the image database system, 2 is an edge line input unit at the time of search, and 3 is the input edge line, for example, the edge periphery as shown in A and B of FIG. An edge peripheral area creating unit 4 for creating an area, a color input unit for a searcher to input a color of the area for each area created by the edge peripheral area creating unit 3,
Reference numeral 5 is a search result display unit.

In the image database system 1, reference numeral 11 is an image storage unit in which image data is stored, 1
Reference numeral 2 denotes an edge image storage unit in which edge image data created by previously performing edge extraction processing on each image stored in the image storage unit 11 is stored. Reference numeral 13 denotes an edge peripheral area for creating an area corresponding to the edge surrounding area created by the edge surrounding area creating unit 3 (this is an edge surrounding area candidate) from the accumulated original image and edge image. This is a candidate creation unit.

Reference numeral 14 is a distance value calculation unit for calculating the distance value between the edge peripheral region and the edge peripheral region candidate, and reference numeral 15 is an example calculated for each accumulated image by the distance value calculation unit 14. It is a display image selection unit that selects an accumulated image having a short distance value (short distance) from the image.

In the system according to the present embodiment having the configuration shown in FIG. 5, at the time of image storage, an image for storage is input to the image storage unit 11 and at the same time, a known edge is added to this image. The edge image obtained by performing the detection process is input to the edge image storage unit 12. That is, the image database system 11 stores the original images and the edge images corresponding to the original images. Note that the edge image is binarized by a known method and stored.

In this embodiment, as an image search method from the image database, the position of the edge and the edge are defined as an example image in which the searcher specifies the position of the edge part of the image stored in the memory and the color around the edge part. Use the color of the surrounding area.

Therefore, a search process for searching an image stored in the image database in this embodiment will be described with reference to the flowchart shown in FIG.

First, in step S1 of FIG. 6, the user draws a polygonal line on the screen as an exemplary image using a known graphic editor or the like. This is because, for example, when a searcher wants to search for an image as shown in FIG. 7, the searcher shows a clear edge in the image by the symbol ABCD in FIG. 8 based on the memory. Specify with a polygonal line so that If the edge of the original image is close to a straight line, it may be designated by a straight line (this polygonal line is hereinafter referred to as an edge line).

In step S2, image expansion processing of x dots (x is a predetermined value) is performed on the edge line designated in step S1 to obtain an area E as shown in FIG.
Next, a segment that passes through both end points (A and D in FIG. 8) of the input edge line and includes both end points (here, AB and CD)
The area E is divided by a straight line perpendicular to, and a portion including the designated edge line is defined as an edge peripheral area among the divided partial areas.

As a result, the edge peripheral region is divided into two regions X and Y by the edge line as shown in FIG. However, when the input start point (here, A) of the edge line and the input end point (here, D) are seen along the edge line, the left side of the two areas is set as X.

In step S3, the edge peripheral area obtained in step S2 is displayed to the searcher. Then, the searcher inputs the color of the portion corresponding to the area of the original image in one or both of the two areas based on the memory of the searcher.

In the subsequent steps S4 to S10 (the process indicated by reference character A in FIG. 6), the process described below is performed for each image stored in the image database.

That is, in step S4, only the portion corresponding to the edge peripheral region obtained in step S2 is extracted from the edge image stored in the edge image storage section 12. For example, if there is an edge image as shown in FIG. 11B for an edge peripheral area having a shape as shown in FIG. 11A, the area shown in FIG. 11A and the image shown in FIG. Of the obtained edges ((c) of FIG. 11), FIG.
In (d) of 1, the portion indicated by the solid line is extracted. Note that, hereinafter, the portion extracted here will be referred to as a corresponding edge candidate.

In step S5 of FIG. 6, the extracted corresponding edge candidate is subjected to binary image expansion processing by y dots (y is a predetermined value) as shown in (e) of FIG. Of the candidates, those with a short distance are integrated. Then, in step S6, the region subjected to the expansion process in step S5 is subjected to a thinning process by a known method, so that this region is connected to a certain extent as shown in (f) of FIG. It is a polygonal line with a branch of 1.

In step S7, the broken line obtained in step S6 is traced in the same manner as a chain code is created, and the longest chain is taken out as shown in FIG. Then, a polygonal line approximation is performed on the chain by the piecewise linear approximation method so that the number of segments included in the input polygonal line is three (three in the example of FIG. 8). The piecewise linear approximation method is described in, for example, Hideyuki Tamura: “Introduction to Computer Image Processing”, Soken Shuppan, pp.93-94.

An example of the result obtained by performing the above polygonal line approximation is shown in FIG. The polygonal line thus obtained is a candidate for the edge corresponding to the input edge line. Note that this is hereinafter referred to as a candidate edge line.

Regarding both end points of the candidate edge line, the one closer to the input start point of the input edge line corresponds to the input start point. In this case, the other end point of the candidate edge line corresponds to the end point (D ′) of the input edge line.

In step S8, two edge peripheral areas are created from the candidate edge line obtained in step S7, as shown in FIG. 11 (i), in the same manner as in step S2. These are hereinafter referred to as edge peripheral areas. Then, of the two end points of the candidate edge line, the other end point (D ′) is seen from the end point (A ′) corresponding to the start point (A) of the input edge line, and the edge peripheral region candidate on the left side thereof is selected. Let X ′ be the edge peripheral region candidate on the right side be Y ′. X'and Y'correspond to X and Y, respectively.

In step S9, with respect to the edge peripheral area candidates obtained in the processing up to step S8, a representative color of the accumulated original image in the area is obtained. There are various methods for obtaining the representative color of a certain area, but it is also possible to simply take the average of the RGB values of all the pixels in the area.

In step S10, the distance between the color input by the searcher in the edge surrounding area in step S3 and the representative color of the edge surrounding area candidate corresponding to the area is calculated. There are several ways to calculate the color distance, but simply RGB
All you have to do is find the distance between the two colors in space. If the color is input to only one of the edge peripheral areas in step S3, the similarity is calculated only for that area and the edge peripheral area candidates corresponding to the area.

When a color is input to both areas, the similarity is calculated between each edge peripheral area and the edge peripheral area candidate corresponding to each area, and those values are averaged. Here, this value is the distance value between the input value and the accumulated image.

Here, in the processes of steps S4 to S10 described above, the distance value between the input value and each accumulated image is repeatedly obtained for all accumulated images. Then, in step S11, a low value (close distance) is selected from the distance values obtained for each accumulated image, and the image is displayed to the searcher as a search result.

As described above, according to this embodiment,
Even if the image you want to search does not have a uniform color area with a certain size, if there is a clear edge in the image, the searcher will specify that edge as a straight line or a polygonal line, and By searching for an image from the image database with the color-designated example image as the example image, the example image can be easily specified, and the search can be performed quickly.

The present invention is not limited to the above embodiment, and for example, matching is performed between the candidate edge line obtained by the processing in step S7 of FIG. 6 and the input edge line, and the matching is performed. If it is difficult, that is, if the edge lines are not similar to each other, the stored image may be excluded from the search target to perform the search at a higher speed.

As a matching method here, for example, as shown in FIG. 12, when the starting point (A) of the input edge line and the end point (A ') of the candidate edge line corresponding to the point are overlapped with each other. , If the ratio of the distance d between the other end points (D, D ′) of both edge lines and the length of the input edge line is larger than a predetermined value,
A method of removing the image that is the source of the candidate edge line from the search target can be considered.

Further, in the above embodiment, the edge input by the searcher based on the memory about the original image does not have to be a polygonal line. For example, for the image shown in FIG. 3 (the image to be searched), You may input the curve as shown in FIG. In this case, by applying a piecewise linear approximation method to this curve, the curve approximated by a polygonal line is regarded as an input edge line, and the same processing as described above may be performed.

However, in this case, unlike the piecewise linear approximation method in the above embodiment, the process is repeated until the distance between the input curve and the polygonal line becomes shorter than a predetermined value.

The present invention may be applied to a system composed of a plurality of devices or an apparatus composed of a single device. Further, it goes without saying that the present invention can be applied to the case where it is achieved by supplying a program to a system or an apparatus.

[0036]

As described above, according to the present invention,
An image search can be performed quickly and reliably based on the degree of comparison between the input edge line position and the color designated around the edge line and the accumulated image data.

[0037]

[Brief description of drawings]

FIG. 1 is a diagram illustrating an example of an image that is stored in a searcher and is about to be searched.

FIG. 2 is a diagram showing, as an exemplary image, a cluster of color-matched regions in FIG.

FIG. 3 is a diagram showing an example of an image in which there is no cluster of color-matched areas in the image.

FIG. 4 is a diagram illustrating an example in which a searcher presents an edge line and colors around the edge line as an exemplary image.

FIG. 5 is a block diagram showing a configuration of an image database system according to an embodiment of the present invention.

FIG. 6 is a flowchart showing an image search processing procedure in an image database according to the present invention.

FIG. 7 is a diagram showing an example of an image that a searcher is searching for according to an embodiment.

FIG. 8 is a diagram showing an example of an edge line presented by a searcher for searching the image shown in FIG. 7.

FIG. 9 is a diagram in which an image expansion process is performed on the edge line presented in FIG.

10 is a diagram showing an example in which an edge peripheral region is generated from the edge line shown in FIG.

FIG. 11 is a diagram showing a flow of image processing for obtaining a distance value between an edge line and a peripheral region thereof based on the edge line.

FIG. 12 is a diagram showing a method of matching an input edge line and a candidate edge line.

FIG. 13 is a diagram illustrating an example in which a portion corresponding to an edge of an original image is input as a curve.

FIG. 14 is a diagram showing an example in which a curved edge is approximated by a polygonal line.

[Explanation of symbols]

1 Image database system 2 Edge line input section 3 Edge peripheral area creation unit 4 color input section 5 Display 11 Image storage unit 12 Edge image storage unit 13 Edge peripheral area candidate creation unit 14 Distance value calculator 15 Display image selection section

─────────────────────────────────────────────────── ─── Continuation of the front page (58) Fields surveyed (Int.Cl. ⁷ , DB name) G06F 17/30 G06T 1/00 G06T 7/00 JISST file (JOIS)

Claims (10)

(57) [Claims] 1. An image retrieval apparatus for retrieving a desired image from a plurality of image data stored in a database, means for inputting an edge line of an exemplary image of an image to be retrieved, and a color for a peripheral region of the edge line. Means for calculating the similarity between the exemplary image and the plurality of image data from the color of the edge line and the peripheral area, and an image corresponding to the image data having the maximum similarity. An image search apparatus comprising: a unit for displaying a searched image. 2. The image search according to claim 1, further comprising calculation means for calculating a distance value between the exemplary image and the plurality of image data, and the distance value is used as the similarity. apparatus. 3. The calculating means further comprises means for accumulating edge images corresponding to the plurality of image data, and means for extracting an edge of a portion corresponding to the edge line from the edge image, The image search device according to claim 2, wherein a distance between the color of the extracted peripheral region of the edge and the designated color is set as the distance value. 4. The image search device according to claim 1, wherein the edge line is a polygonal line. 5. The image search device according to claim 1, wherein the edge line is a curve. 6. An image retrieval method for retrieving a desired image from a plurality of image data stored in a database, the step of inputting an edge line of an exemplary image of an image to be retrieved, and a color for a peripheral region of the edge line. And a step of calculating the degree of similarity between the example image and the plurality of image data from the edge line and the color of the peripheral area, and the image corresponding to the image data with the highest degree of similarity And a step of displaying the searched image as an image to be searched. 7. The image search method according to claim 6, further comprising a step of calculating a distance value between the exemplary image and the plurality of image data, and the distance value is used as the similarity. . 8. The calculating step further comprises a step of accumulating edge images corresponding to the plurality of image data, and a step of extracting an edge of a portion corresponding to the edge line from the edge image, The image search method according to claim 7, wherein the distance between the color of the peripheral region of the extracted edge and the designated color is set as the distance value. 9. The image search method according to claim 6, wherein the edge line is a polygonal line. 10. The image search method according to claim 6, wherein the edge line is a curve.