IMAGE RETRIEVAL SYSTEM AND METHOD

• US 20120230610A1
• Filed: 05/21/2012
• Published: 09/13/2012
• Est. Priority Date: 01/12/2009
• Status: Active Grant

• Current Assignee: Gold Charm Limited (Hon Hai Precision Industry Co., Ltd.)
• Original Assignee: Hong Fu Jin Precision Industry (Shenzhen) Co., Ltd. (Hon Hai Precision Industry Co., Ltd.), Hon Hai Precision Industry Co., Ltd.
• Inventors: LEE, CHUNG-I, YEH, CHIEN-FA, XIAO, WEI-QING
• Granted Patent: US 8,913,853 B2
• Time in Patent Office: Days
• US Class Current: 382/305
• CPC Class Codes: G06F 16/50 of still image data
• CPC Class Codes: G06F 16/53 Querying
• CPC Class Codes: G06F 16/5838 using colour

Abstract

An image retrieval method applies an application server, one or more calculating servers, and a sorting server to perform image retrieval. The application server extracts visual features of an exemplary image. The one or more calculating servers calculate similarities of available images according to the visual features of the exemplary image. The sorting server sorts the available images according to the similarities so as to obtain images similar to the exemplary image.

Claims

1. An image retrieval system for searching images similar to an exemplary image from available images, the image retrieval system comprising:

a storage system;

at least one processor;

an image retrieval unit comprising one or more computerized codes that are stored in the storage system and executed by the at least one processor, the one or more computerized codes comprising;

an extracting module operable to extract visual features of the exemplary image and transfer the visual features of the exemplary image to one or more calculating servers, wherein the visual features of the exemplary image comprise color features and shape features, and wherein the visual features of the exemplary image are extracted by dividing the exemplary image into a plurality of image blocks, calculating an arithmetic mean of gray values of pixels in each image block as color features of the image block, calculating a ratio of black pixels to total pixels in the image block as shape features of the image block, and determining overall visual features of the exemplary image according to the color features and the shape features of the image blocks;

a distributing module operable to allocate image comparison tasks of the available images to the one or more calculating servers, such that the one or more calculating servers calculate similarities of the available images to the exemplary image according to the visual features of the exemplary image and predetermined visual features of the available images, fetch image indexes of the available images from an index server, and return the similarities and the image indexes of the available images to the image retrieval system;

a gathering module operable to gather the similarities and the image indexes of all the available images from the one or more calculating servers;

a sorting module operable to transfer the gathered similarities of the available images to a sorting server, such that the sorting server sorts the available images according to the gathered similarities of the available images; and

an outputting module operable to receive a sorting sequence of the available images from the sorting server, and output the image indexes of the available images in the sorting sequence to a client computer.

2. The image retrieval system of claim 1, wherein each of the available images is divided into a same amount of image blocks as the exemplary image, a similarity of each image block of an available image is calculated, and an overall similarity of the available image to the exemplary image is calculated according to the similarities of all the image blocks of the available image.

3. The image retrieval system of claim 2, wherein an arithmetic mean of the similarities of all the image blocks of the available image is calculated as the overall similarity of the available image to the exemplary image.

4. The image retrieval system of claim 2, wherein the similarity of an image block of the available image is determined according to a difference between the image block of the available image and a corresponding image block of the exemplary image.

5. The image retrieval system of claim 1, wherein the image indexes of the available images comprise image locations, metadata, and visual features of the available images.

6. The image retrieval system of claim 5, wherein the metadata of the available images comprise captions, keywords, and/or descriptions of the available images.

7. The image retrieval system of claim 1, wherein the exemplary image and the available images are in portable document format (PDF), joint photographic experts group (JPEG), graphic interchange format (GIF), or tagged image file format (TIFF).

8. The image retrieval system of claim 1, wherein the image comparison tasks are equally allocated to the calculating servers.

9. The image retrieval system of claim 1, wherein a specified amount of the image indexes of the available images is output.

10. A computer-based image retrieval method for searching images similar to an exemplary image from available images, the image retrieval method being executed by a processor of a computing device and comprising:

extracting visual features of the exemplary image and transferring the visual features of the exemplary image to one or more calculating servers, wherein the visual features of the exemplary image comprise color features, and the color features of the exemplary image are extracted by dividing the exemplary image into a plurality of image blocks, calculating an arithmetic mean of gray values of pixels in each image block as color features of the image block, and determining overall color features of the exemplary image according to the color features of the image blocks;

distributing image comparison tasks of the available images to the one or more calculating servers, such that the one or more calculating servers calculate similarities of the available images to the exemplary image according to the visual features of the exemplary image and predetermined visual features of the available images, fetch image indexes of the available images from an index server, and returns the similarities and the image indexes of the available images;

gathering the similarities and the image indexes of all the available images from the one or more calculating servers;

transferring the gathered similarities of the available images to a sorting server, and causing the sorting server to sort the available images according to the gathered similarities of the available images; and

receiving a sorting sequence of the available images from the sorting server, and outputting the image indexes of the available images in the sorting sequence to a client computer.

11. The image retrieval method of claim 10, wherein each of the available images is divided into a same amount of image blocks as the exemplary image, and a similarity of one available image is calculated by calculating a similarity of each image block of the one available image, and calculating an overall similarity of the one available image according to the similarities of all the image blocks of the one available image.

12. The image retrieval method of claim 11, wherein an arithmetic mean of the similarities of all the image blocks of the available image is calculated as the overall similarity of the available image to the exemplary image.

13. The image retrieval method of claim 11, wherein the similarity of an image block of the available image is determined according to a difference between the image block of the available image and a corresponding image block of the exemplary image.

14. A computer-based image retrieval method for searching images similar to an exemplary image from available images, the image retrieval method being executed by a processor of a computing device and comprising:

extracting visual features of the exemplary image and transferring the visual features of the exemplary image to one or more calculating servers, wherein the visual features of the exemplary image comprise shape features, and the shape features of the exemplary image are extracted by dividing the exemplary image into a plurality of image blocks, calculating a ratio of black pixels to total pixels in the image block as shape features of the image block, and determining overall shape features of the exemplary image according to the shape features of the image blocks;

distributing image comparison tasks of the available images to the one or more calculating servers, such that the one or more calculating servers calculate similarities of the available images to the exemplary image according to the visual features of the exemplary image and predetermined visual features of the available images, fetch image indexes of the available images from an index server, and returns the similarities and the image indexes of the available images;

gathering the similarities and the image indexes of all the available images from the one or more calculating servers;

transferring the gathered similarities of the available images to a sorting server, and causing the sorting server to sort the available images according to the gathered similarities of the available images; and

receiving a sorting sequence of the available images from the sorting server, and outputting the image indexes of the available images in the sorting sequence to a client computer.

15. The image retrieval method of claim 14, wherein each of the available images is divided into a same amount of image blocks as the exemplary image, and a similarity of one available image is calculated by calculating a similarity of each image block of the one available image, and calculating an overall similarity of the one available image according to the similarities of all the image blocks of the one available image.

16. The image retrieval method of claim 15, wherein an arithmetic mean of the similarities of all the image blocks of the available image is calculated as the overall similarity of the available image to the exemplary image.

17. The image retrieval method of claim 15, wherein the similarity of an image block of the available image is determined according to a difference between the image block of the available image and a corresponding image block of the exemplary image.

18. A computer-based image retrieval method for searching images similar to an exemplary image from available images, the image retrieval method being executed by a processor of a computing device and comprising:

extracting visual features of the exemplary image and transferring the visual features of the exemplary image to one or more calculating servers, wherein the exemplary image is divided into a plurality of image blocks;

distributing image comparison tasks of the available images to the one or more calculating servers, such that the one or more calculating servers calculate similarities of the available images to the exemplary image according to the visual features of the exemplary image and predetermined visual features of the available images, fetch image indexes of the available images from an index server, and returns the similarities and the image indexes of the available images, wherein each of the available images is divided into a same amount of image blocks as the exemplary image, and a similarity of one available image is calculated by calculating a similarity of each image block of the one available image, and calculating an overall similarity of the one available image according to the similarities of all the image blocks of the one available image;

gathering the similarities and the image indexes of all the available images from the one or more calculating servers;

transferring the gathered similarities of the available images to a sorting server, and causing the sorting server to sort the available images according to the gathered similarities of the available images; and

receiving a sorting sequence of the available images from the sorting server, and outputting the image indexes of the available images in the sorting sequence to a client computer.

19. The image retrieval method of claim 18, wherein the visual features of the exemplary image comprise color features and shape features.

20. The image retrieval method of claim 19, wherein an arithmetic mean of gray values of pixels in each image block is calculated as color features of the image block, a ratio of black pixels to total pixels in the image block is calculated as shape features of the image block, and overall visual features of the exemplary image is determined according to the color features and the shape features of the image blocks.

21. The image retrieval method of claim 18, wherein an arithmetic mean of the similarities of all the image blocks of the available image is calculated as the overall similarity of the available image to the exemplary image.

22. The image retrieval method of claim 18, wherein the similarity of an image block of the available image is determined according to a difference between the image block of the available image and a corresponding image block of the exemplary image.

23. The image retrieval method of claim 18, wherein the image indexes of the available images comprise image locations, metadata, and visual features of the available images.

24. The image retrieval method of claim 23, wherein the metadata of the available images comprise captions, keywords, and/or descriptions of the available images.

25. The image retrieval method of claim 18, wherein the exemplary image and the available images are in portable document format (PDF), joint photographic experts group (JPEG), graphic interchange format (GIF), or tagged image file format (TIFF).

26. The image retrieval method of claim 18, wherein the image comparison tasks are equally allocated to the calculating servers.

27. The image retrieval method of claim 18, wherein a specified amount of the image indexes of the available images is output.

28. A computer-based image retrieval method for searching images similar to an exemplary image from available images, the image retrieval method being executed by a processor of a computing device and comprising:

extracting visual features of the exemplary image, wherein the visual features of the exemplary image comprise color features, and the color features of the exemplary image are extracted by dividing the exemplary image into a plurality of image blocks, calculating an arithmetic mean of gray values of pixels in each image block as color features of the image block, and determining overall color features of the exemplary image according to the color features of the image blocks;

calculating similarities of the available images to the exemplary image according to the visual features of the exemplary image and predetermined visual features of the available images, and fetching image indexes of the available images from an index server;

sorting the available images according to the similarities of the available images to obtain a sorting sequence of the available image; and

outputting the image indexes of the available images in the sorting sequence to a client computer.

29. A computer-based image retrieval method for searching images similar to an exemplary image from available images, the image retrieval method being executed by a processor of a computing device and comprising:

extracting visual features of the exemplary image, wherein the visual features of the exemplary image comprise shape features, and the shape features of the exemplary image are extracted by dividing the exemplary image into a plurality of image blocks, calculating a ratio of black pixels to total pixels in the image block as shape features of the image block, and determining overall shape features of the exemplary image according to the shape features of the image blocks;

calculating similarities of the available images to the exemplary image according to the visual features of the exemplary image and predetermined visual features of the available images, and fetching image indexes of the available images from an index server;

sorting the available images according to the similarities of the available images to obtain a sorting sequence of the available image; and

outputting the image indexes of the available images in the sorting sequence to a client computer.

30. A computer-based image retrieval method for searching images similar to an exemplary image from available images, the image retrieval method being executed by a processor of a computing device and comprising:

extracting visual features of the exemplary image, wherein the exemplary image is divided into a plurality of image blocks;

calculating similarities of the available images to the exemplary image according to the visual features of the exemplary image and predetermined visual features of the available images, and fetching image indexes of the available images from an index server, wherein each of the available images is divided into a same amount of image blocks as the exemplary image, and a similarity of one available image is calculated by calculating a similarity of each image block of the one available image, and calculating an overall similarity of the one available image according to the similarities of all the image blocks of the one available image;

sorting the available images according to the similarities of the available images to obtain a sorting sequence of the available image; and

outputting the image indexes of the available images in the sorting sequence to a client computer.

31. The image retrieval method of claim 30, wherein the visual features of the exemplary image comprise color features and shape features.

32. The image retrieval method of claim 31, wherein an arithmetic mean of gray values of pixels in each image block is calculated as color features of the image block, a ratio of black pixels to total pixels in the image block is calculated as shape features of the image block, and overall visual features of the exemplary image is determined according to the color features and the shape features of the image blocks.

33. The image retrieval method of claim 30, wherein an arithmetic mean of the similarities of all the image blocks of the available image is calculated as the overall similarity of the available image to the exemplary image.

34. The image retrieval method of claim 30, wherein the similarity of an image block of the available image is determined according to a difference between the image block of the available image and a corresponding image block of the exemplary image.

35. The image retrieval method of claim 30, wherein the image indexes of the available images comprise image locations, metadata, and visual features of the available images.

36. The image retrieval method of claim 30, wherein the metadata of the available images comprise captions, keywords, and/or descriptions of the available images.

37. The image retrieval method of claim 30, wherein the exemplary image and the available images are in portable document format (PDF), joint photographic experts group (JPEG), graphic interchange format (GIF), or tagged image file format (TIFF).

38. The image retrieval method of claim 30, wherein a specified amount of the image indexes of the available images is output.