JPH05266091A - Similar color tone retrieving device for image - Google Patents

Abstract

PURPOSE:To judge the similarity of color tones between images without interposing any judge person by calculating the similarity of key color vectors between images and judging the similarity of color tones between images by using the degree of similarity for key color vectors. CONSTITUTION:The key color vectors are added to image data stored in an image storage part 2. When the retrieval instruction of an image having the color tone similar to that of a certain image is transmitted from a control part 1, the key color vectors of images are successively extracted from the image storage part 2, and the similarity between the respective extracted vectors and the key color vector of the image to be retrieved is judged by a similar color tone judge part 5 to judge it corresponding the calculated result of cosine similarity. Then, the image having the vector judged similar to the key color vector of the image to be retrieved by the judged result is displayed at a display part 3 as the similar color tone image.

Description

Detailed Description of the Invention

[0001]

BACKGROUND OF THE INVENTION 1. Field of the Invention The present invention relates to an image search device.

[0002]

2. Description of the Related Art FIG. 2 is a block diagram showing, for example, a conventional similar image tone retrieving apparatus. 1 is a control unit for controlling the entire retrieving apparatus, 2 is an image storage unit for recording and reproducing image data, and 3 is an image. The display unit 4 for displaying the images stored in the storage unit 2 is a judge who determines whether or not the color tones of the images are similar.

Next, the operation will be described. In the conventional image search system, when searching for an image having a color tone similar to this image for a certain image, for example, the searched image and the image stored in the storage unit 2 are simultaneously displayed and displayed. The search is performed with the intervention of the determiner, such as determining whether the determined image has a similar color tone by the determiner 4, and performing this operation for all the images stored in the image storage unit 2.

[0004]

Since the conventional image similar color tone retrieving apparatus is configured as described above, the judge must judge whether or not there is a similarity in color tone between images. There is a problem in that it takes time and labor, and variations occur depending on different judges.

The present invention has been made to solve the above problems, and an object thereof is to obtain an image similar color tone retrieval device capable of determining the similarity of the color tone between images without the intervention of a determiner. To do.

[0006]

An image similar color tone retrieving apparatus according to the present invention defines a key color vector representing a color tone of an image, assigns this key color vector to each image, and sets a key color vector between images. And the similarity of the color tones of the image can be determined by using the degree of similarity of the key color vector.

The similarity of key color vectors between images is
Cosine similarity measure, set-theoretic similarity measure and distance are used.
Each element of the key color vector indicates whether or not the key color representing the color tone is included in the image and the degree of the inclusion.

[0008]

In the key color vector given to the image in the present invention, the similarity of the color tone between the images is shown by calculating the similarity between the key color vectors, and the similar color tone of the image is judged.

[0009]

【Example】

Example 1. FIG. 1 is a block diagram showing a first embodiment of the present invention, in which 1 to 3 are the same as those of the conventional device, but a key color vector is added to the image data stored in the image storage unit 2. There is. With the key color vector, the color representing the image is divided into several types, and the representative color is used as the key color, and the element corresponding to the key color is 1 or 0 depending on whether or not each key color is included in the image. That is, a set of images is P (| P | = N) and a set of key colors is R (|
When R | = M), the key color vector of the image p _i εP is represented by the following equation. v _pi = (δ _i ¹ , δ _i ² , ..., δ _i ^M ) i = 1, 2, ..., N However, when the image p _i has the key color r _u (εR), δ _i ^u = 1 When the image p _i does not have the key color r ^u (∈R), δ _i ^u = 0

Reference numeral 5 is a similar color tone determination unit that determines the similarity of the color tone between the images from the key color vector added to the images.

Next, the operation will be described. In the image search system, when a search command for an image having a color tone similar to that of a certain image is sent from the control unit 1, key color vectors of the image are sequentially fetched from the image storage unit 2 and the fetched vectors and the target color vector are fetched. The similarity of the key color vector of the search image is determined by the similar color tone determination unit 5, and an image having a vector which is determined to be a similar vector to the key color vector of the searched image based on the determination result is displayed as the similar color tone image on the display unit 3
To display.

The similar color tone determination section 5 first calculates the cosine similarity measure s of the key color vector v _p0 of the searched image and the key color vector v _pi of the image p _i . The cosine similarity measure is calculated by the following equation 1.

[0013]

[Equation 1]

Next, if the similarity measure s is greater than or equal to the threshold value, it is determined that the color tones of these vectors, that is, the images having these vectors are similar.

Example 2. Although the similarity measure s is calculated using the cosine similarity measure in the first embodiment, the set theory similarity measure is used in the second embodiment. The set-theoretic similarity measure s (x, y) of vector x and vector y is shown below. s (x, y) = E (U _x ∩U _y ) / E (U _x ∪U _y ), where U _x is a set of components whose vector x is one, and E (U _x ∩U _y ). Is the number of elements of the intersection of U _x and U _y , E (U _x
∪ U _y ) is the number of elements in the union of U _x and U _y . The use of this set-theoretic similarity measure facilitates the calculation of the similarity measure.

Embodiment 3. In the first and second embodiments, the similarity measure is used to determine the similarity of the image color tones, but in the third embodiment, the distance between the key color vectors is used. The Hamming distance is used as the distance between the vector x and the vector y because its element is 0 or 1. The Hamming distance is shown in Equation 2 below.

[0017]

[Equation 2]

It is assumed that the shorter the distance is, the more similar the vectors are, that is, the images are.

By using the distance to determine the similarity, the following triangular inequality holds, so that the similarity relationship more closely matches the actual feeling. d (x, y) + d (y, z) ≥ d (x, y)

Example 4. In the first embodiment, the element of the key color vector is set to 1 or 0 depending on whether or not the key color is included, but in the fourth embodiment, the key has a weight depending on the degree to which the key color is included. Construct a color vector. Therefore, the key color vector has the following configuration.

V _pi = (ω _i ¹ , ω _i ² , ..., ω _i ^M ) i = 1, 2, ..., N, where ω _i ^u has an image p _i with a key color r _u degree

By expressing the elements of the key color vector by weights, the color tone of the image can be expressed more finely.

By the way, it goes without saying that the similarity measure may be replaced with the distance even when the key color vector is represented by the weight. In this case, the Euclidean distance shown in the following Expression 3 is used as the distance.

[0024]

[Equation 3]

[0025]

Since the present invention is constructed as described above, it has the following effects.

Since the color tone of the image is formed as the key color vector and the similarity of the key color vector is determined by the calculation result of the cosine similarity measure, the similarity of the color tone between the images can be determined without the intervention of the determiner.

Also, by using the set-theoretic similarity measure as the similarity measure, the apparatus for calculating the similarity measure is simplified and the calculation speed, that is, the judgment speed is improved.

Further, by using the distance between vectors as the similarity, it is possible to propose a similarity that is more realistic.

Further, by using the key color vector as the weight expression, the color tone of the image can be expressed in detail and the similarity can be determined more finely.

[Brief description of drawings]

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

FIG. 2 is a block diagram of a conventional image similar color tone search device.

[Explanation of symbols]

 1 control unit 2 image storage unit 3 display unit 5 similar color tone determination unit

─────────────────────────────────────────────────── ───

[Procedure amendment]

[Submission date] June 8, 1992

[Procedure Amendment 1]

[Document name to be amended] Statement

[Correction target item name] 0013

[Correction method] Change

[Correction content]

[0013]

[Equation 1]

Claims (4)

[Claims] 1. An image storage unit for storing and reproducing image data to which a key color vector having a code indicating whether or not a certain color is included in the image is added, a display unit for displaying the image data, and image similarity. An image similar tone retrieving apparatus, comprising: a similarity determination unit that calculates a sex using a cosine similarity measure of a key color vector; and a control unit that controls the similarity determination unit. 2. The image similar color tone retrieving apparatus according to claim 1, further comprising means for calculating the image similarity by using a set-theoretic similarity measure of the key color vector. 3. The image similar color tone retrieving apparatus according to claim 1, further comprising means for calculating similarity of images by using a distance between key color vectors. 4. The image similar tone retrieving apparatus according to claim 1, further comprising means for storing image data to which a key color vector having a degree of a color included in the image as a weight is added.