Method for retrieving image by using multiple features per image subregion

• US 6,658,168 B1
• Filed: 05/30/2000
• Issued: 12/02/2003
• Est. Priority Date: 05/29/1999
• Status: Expired due to Term

• Current Assignee: LG Electronics, Inc. (LG Corporation)
• Original Assignee: LG Electronics, Inc. (LG Corporation)
• Inventors: Kim, Hyeon Jun
• Primary Examiner(s): Wu, Jingge
• Application Number:
  US09/583,221
• Time in Patent Office: 1,281 Days
• Field of Search: 382/162-167, 382/305, 707/1-104
• US Class Current: 382/305
• CPC Class Codes: G06V 10/507 Summing image-intensity values; Histogram projection analysis
• CPC Class Codes: Y10S 707/99945 Object-oriented database structure processing
• CPC Class Codes: Y10S 707/99948 Application of database or data structure, e.g. distributed, multimedia, or image

Abstract

An efficient method for retrieving an image using multiple features for an image subregion is disclosed. The present invention obtains a regional representative color for the subregions and uses the regional representative color during a retrieval of a similar image if the regional representative color is reliable. Otherwise, a feature information other than the regional representative color is used.

Claims

1. A method for retrieving an image comprising:

dividing a reference image into subregions, extracting a regional representative color and a confidence measure of the regional representative color for each of the subregions, and extracting a second feature information of the reference image;

comparing the confidence measure of the regional representative color for each subregion to a predetermined threshold value; and

determining a similarity between the reference image and a target image using the regional representative color if the confidence measure of the regional representative color for a subregion is greater than the threshold value, and otherwise, determining a similarity between the reference image and a target image using the second feature information of the reference image.

2. A method of claim 1, wherein the reference image is divided into subregions of any arbitrary form.

3. A method of claim 1, wherein the reference image is divided by a method based upon an object.

4. A method of claim 1, wherein the reference image is divided into grid regions.

5. A method of claim 4, wherein the regional representative color is an average color obtained by an average filtering of a grid of a fixed size.

6. A method of claim 4, wherein the second feature information of the reference image is one of or a combination of both a plurality of main colors of the grids and a frequency of each main color, and a grid feature information other than a color information.

7. A method of claim 6, wherein the plurality of main colors of the grids and a frequency of each main color ate obtained from a histogram.

8. A method of claim 6, wherein the grid feature information other than a color information is a texture information.

9. A method of claim 8, wherein the texture information is a edge direction components.

10. A method of claim 6, wherein the regional representative color of a subregion for the reference image is obtained from the plurality of main colors of the grids and a frequency of each main color at a time during a retrieval of the target image.

11. A method of claim 1, wherein the second feature information of the reference image is one of or a combination of both a plurality of main colors of the subregions and a frequency of each main color, and a subregion feature information other than a color information.

12. A method of claim 11, wherein the plurality of main colors of the subregions and a frequency of each main color are obtained from a histogram.

13. A method of claim 6, wherein the subregion feature information other than a color information is a texture information.

14. A method of claim 13, wherein the texture information is a edge direction components.

15. A method of claim 11, wherein the regional representative color of a subregion for the reference image is obtained from the plurality of main colors of the subregions and a frequency of each main color at a time during a retrieval of the target image.

16. A method for retrieving an image in an image retrieval system comprising:

dividing a reference image into grid regions;

extracting a regional representative color and a confidence measure of the regional representative color for each grid region; and

performing a similarity examination between the reference image and a target image using the regional representative color and the confidence measure of the regional representative color for each grid region, wherein the performing a similarity examination uses the regional representative color if said confidence measure of the regional representative color is greater than a predetermined threshold value, and otherwise the performing a similarity-examination uses a feature information of the reference image other than the regional representative color.

17. A method of claim 16, wherein the feature information of the reference image other than the regional representative color is a color histogram and a texture information.

18. A method of claim 16, wherein the feature information of the reference image other than the regional representative color is a color histogram.

19. A method of claim 16, wherein the feature information of the reference image other than the regional representative color is a texture information.