JPH10334117A - Computer-readable recording medium where similar image retrieval program is recorded - Google Patents

Abstract

PROBLEM TO BE SOLVED: To provide the computer-readable recording medium which enables image retrieval without key word input and has the similar image retrieval program recorded. SOLUTION: Coordinate information and color information are normalized as to respective source image data of many retrieval images stored in the same recording medium and an example image that a user inputs to a computer. Then the similarities of the respective normalized retrieval image data to the normalized example image data are found. Further, the retrieval images are sequenced in the decreasing order of the similarities and high-order retrieval images are outputted.

Description

DETAILED DESCRIPTION OF THE INVENTION

[0001]

BACKGROUND OF THE INVENTION 1. Field of the Invention The present invention relates to a similar image search program capable of storing a large number of images in a recording medium and searching for a required image from the stored images without requiring a keyword input. The present invention relates to a recorded computer-readable recording medium.

[0002]

2. Description of the Related Art Conventionally, there is an image database CD-ROM in which materials for CG creation such as landscape photographs, wood grain patterns, marble patterns, and geometric patterns are collected. In these, there is no search means using only image data, or a search keyword is attached to image data in advance, and a search means for extracting image data to which a keyword to be collated with input of the keyword is added is attached. there were.

[0003]

However, if there is no search means using only image data, it is inconvenient to search for a necessary image from the images because the number of images is enormous.

[0004] In the keyword search, it is necessary to assign a search keyword to each image in advance.
It takes time to enter keywords. Also, it was very difficult to unify keywords.

Further, it is very difficult to distinguish patterns such as wood grain and stone grain by using keywords.

SUMMARY OF THE INVENTION Accordingly, an object of the present invention is to solve the above-mentioned problems, and to provide a computer-readable recording medium storing a similar image search program capable of searching for an image without requiring a keyword input. To provide.

[0007]

In order to achieve the above-mentioned object, a computer-readable recording medium of the present invention is provided with a large number of search images stored in the same recording medium and input to a computer by a user. A procedure for normalizing coordinate information and color information for each original image data of the exemplified image, a procedure for obtaining a similarity to the normalized example image data for each normalized search image data,
A program for causing a computer to execute a procedure for ranking search images in descending order of similarity and a procedure for outputting a higher-order search image is recorded.

Further, in the above configuration, the normalized search image data is stored in advance in the recording medium together with the original image data of the search image, and the procedure for normalizing the original image data of the search image is omitted. Good.

In the above configuration, a large number of search images stored in the recording medium may include an image recorded by a user.

[0010]

DETAILED DESCRIPTION OF THE PREFERRED EMBODIMENTS Hereinafter, a computer-readable recording medium on which a similar image search program of the present invention is recorded will be described in more detail with reference to the drawings.

FIG. 1 is a diagram showing an embodiment of a similar image retrieval system using a recording medium according to the present invention. In FIG. 1, reference numeral 1 denotes an input device for inputting an exemplary image as a search key to a computer, and an optical reading device such as a scanner or a pointing device such as a mouse or a pen tablet is used. Reference numeral 2 denotes a recording medium of the present invention in which a large number of search images and similar image search programs are recorded, and a CD-ROM, a DVD-ROM, an MO, or the like can be used. 3 is a computer main body, and 4 is
Is a storage device, and 5 is an arithmetic device. Reference numeral 6 denotes an output device for displaying a search result, and a display such as a CRT or an LED is used.

[0012] Each original image data of a large number of search images stored in a recording medium and an exemplary image input to a computer by a user is a two-dimensional image, that is, color information (two-dimensional Euclidean space). Any of the following data structures (1) can be used as long as the RGB system is constituted by the following data structure (1) that is discretized.

Original image data (x, y, r, g, b) (1)

The above x and y indicate the discrete points constituting the image, that is, the coordinate position of each pixel, and are each represented by an integer of 0 or more. Also, r, g, and b indicate that the pixel at that coordinate is red (re
Indicates which color among d), green (green), and blue (blue), and the degree of color development (intensity), and is represented by an integer of 0 or more and has the same maximum value for the three colors. Note that the original image data of the search image and the example image preferably have the maximum value of x and y of 20 or more, that is, an image size of a resolution of 21 × 21 = 441 pixels or more. R, g, b
Is preferably 3 or more, that is, 4 × 4 × 4 = 64 gradations or more.

The original image data may be one frame cut from a moving image or two-dimensional data converted from a three-dimensional image.

Further, a large number of search images stored in the recording medium may include an image recorded by a user.

Hereinafter, the similar image retrieval processing according to the present invention will be described with reference to the flowchart of FIG.

First, in step S1 of FIG. 2, coordinate information and color information are normalized for each original image data of a search image and an example image.

The normalization of the coordinate information is to unify the resolutions of the retrieval image and the example image so that they can be compared. The normalization of the color information is to unify the color gradations of the search image and the example image so that they can be compared. When unifying the resolution and the gradation of the search image and the example image, it is desired to unify the resolution and the color gradation at a low resolution. This is because an image having a large number of pixels and color gradations has a large data amount, so that the processing load on subsequent arithmetic processing increases.

The following image data (2) is an example showing a case where the image size is set to 10 × 10 = 100 pixels and the color gradation is set to 4 × 4 × 4 = 64 gradations for the search image and the example image. .

Normalized image data (x, y, r, g, b) x, y = [0, 1, 2,... 9] r, g, b = [0, 1, 2, 3]. 2)

In step S2 in FIG. 2, as shown in the following (3), a difference between each color data on the same coordinate between the normalized example image data and each search image data is calculated to obtain difference data. .

Difference data n (x, y, Δr, Δg, Δb) = retrieval image data n—example image data Δr = rn−r Δg = gn−g Δb = bn−b (3)

In step S3 of FIG. 2, a deviation value is obtained from the absolute value of the difference data. That is, as shown in the following (4), the absolute value | Δr of each difference data for each search image data
|, | Δg |, | Δb |.

Deviation value Dn = Σ | Δr | + Σ | Δg | + Σ | Δb | (4)

The obtained divergence value represents the similarity of each search image data to the example image data, and the smaller the divergence value, the higher the similarity.

In step S4 in FIG. 2, the search images are ranked in descending order of similarity as shown in (5) below.

S1 = min (D0, ..., Dn) S2 = min (D0, ..., Dn) (excluding S1) S3 = min (D0, ..., Dn) (excluding S1, S2) min (D0, ..., Dn) (excluding S1, ..., Sm-1) m = [1, ..., n] ... (5)

Although the search images are similar to the example images in the order of S1, S2, S3,... For example, if up to eight search images similar to the example image are to be obtained, m = 8.

In step S5 of FIG. 2, the higher-order search image is output. At this time, the upper search images may be displayed in a list on the same screen of the display, or the upper search images may be displayed while updating the screen. It should be noted that the search image referred to here includes not only the original image data before being normalized for the search but also image data whose size or the like has been converted for display, for example, for the above list display. After the output, the user specifies and designates a large number of higher-order search images displayed on the display, and the selected search images are printed or used as CG creation materials.

The present invention is not limited to the above embodiment. For example, the search image data normalized in advance together with the original image data of the search image may be stored in the recording medium. In that case, step S in FIG.
1 can omit the normalization of the original image data of the search image.

Further, the similarity of each search image to the example image may be obtained by processes other than those shown in S2 and S3.

In the above description, the xy coordinate system is used as the coordinate information, and the RGB system is used as the color information. However, a polar coordinate system and a CMYK system may be used, respectively.

Further, the recording medium of the present invention may store character and sound information related to the search image together with the search image, and output the image together with the search result.

As a specific example of the similar image search according to the present invention, a clothing designer has a similar image from a product catalog database such as a tie cloth recorded on a recording medium based on images such as photographs and sketches. It is possible to search for. In this case, a sales clerk or a purchaser may use it.

In addition, the following can be considered. (A) Searching for an image having a similar image from an art database recorded on a recording medium based on an image such as a photograph or a sketch. (B) Searching for an image having a similar image from an academic database such as archeology recorded on a recording medium based on an image such as a photograph or a sketch. (C) A case database having similar images is searched from a case database recorded on a recording medium based on images such as medical examination results of a patient. (D) Searching for an image having a similar image from a patent application database recorded on a recording medium based on images such as drawings and sketches. (E) Search for an image having a similar image from a music database recorded on a recording medium based on an image such as a musical score. (F) Based on an image such as a still photograph of a certain scene, a movie having a similar image is searched from a movie frame database recorded on a recording medium.

[0037]

According to the present invention, a computer-readable recording medium having recorded thereon a similar image retrieval program according to the present invention has the following effects.

That is, a procedure for normalizing coordinate information and color information for each of a large number of search images stored in the same recording medium and original image data of an exemplary image input to a computer by a user. A program for causing a computer to execute a procedure for obtaining a similarity to normalized example image data for each search image data, a procedure for ranking search images in descending order of similarity, and a procedure for outputting a higher-order search image is recorded. Therefore, if this is used, an image search can be performed without requiring a keyword input.

[Brief description of the drawings]

FIG. 1 is a diagram showing an embodiment of a similar image search system using a recording medium according to the present invention.

FIG. 2 is a flowchart illustrating an embodiment of a similar image search process according to the present invention.

[Explanation of symbols]

 DESCRIPTION OF SYMBOLS 1 Input device 2 Recording medium 3 Computer main body 4 Storage device 5 Processing device 6 Output device

Claims (3)

[Claims] 1. A procedure for normalizing coordinate information and color information for each of a large number of search images stored in the same recording medium and original image data of an exemplary image input to a computer by a user. A program for causing a computer to execute a procedure for obtaining a similarity to normalized example image data for each search image data, a procedure for ranking search images in descending order of similarity, and a procedure for outputting a higher-order search image is recorded. Computer readable recording medium. 2. The method according to claim 1, wherein the normalized search image data is stored together with the original image data of the search image in a recording medium in advance, and the step of normalizing the original image data of the search image is omitted. Recording medium. 3. A large number of search images stored in a recording medium include an image recorded by a user.
A recording medium according to claim 2.