Image retrieval device, image retrieval method and storage medium storing similar-image retrieval program

• US 20060093221A1
• Filed: 12/14/2005
• Published: 05/04/2006
• Est. Priority Date: 12/10/1999
• Status: Active Grant

• Current Assignee: NEC Corporation
• Original Assignee: NEC Corporation
• Inventors: Kasutani, Eiji
• Granted Patent: US 7,664,333 B2
• Time in Patent Office: Days
• US Class Current: 382/191
• CPC Class Codes: G06F 16/5838 using colour
• CPC Class Codes: G06V 10/431 Frequency domain transformation; Autocorrelation
• CPC Class Codes: Y10S 707/99936 Pattern matching access

Abstract

An image retrieval device and an image retrieval method are provided which are capable of improving performance of image retrieval, of retrieving images at higher speed with simplified configurations and of retrieving images by simplified calculating processes. The image retrieval device is composed of a first coefficient transforming section to transform a first group of image feature descriptors extracted from image data accumulated in an image database and to generate a second group of image feature descriptors to be used for calculating similarity, second coefficient transforming section to transform the first group of image feature descriptors extracted from image data of an inquired image and to generate a second group of image feature descriptors to be used for calculating similarity and a similarity calculating section to compare the second group of image feature descriptors for each piece of image data produced by the coefficient transforming section with the second group of image feature descriptors for the inquired image transformed by the second coefficient transforming section.

Claims

1. An image feature generating device for generating image feature descriptors for similarity calculation from input image data configured so as to perform a DCT (discrete cosine transform) on the input image data to produce DCT coefficients thereon, and perform coefficient transform on the produced DCT coefficients in a manner that fine quantization is carried out on a range of a portion of each of the produced DCT coefficients having a small amplitude and coarse quantization is carried out on a range of a portion of each of the produced DCT coefficients having a large amplitude, in order to generate the image feature descriptors, using the transformed and quantized DCT coefficients as the image feature descriptors.

2. The image feature generating device according to claim 1, configured so as to generate second image data having a specified number of pixels based on input first image data, and perform the DCT (discrete cosine transform) on the generated second image data as the input image data to produce the DCT coefficients.

3. The image feature generating device according to claim 1, configured so as to partition first image data into specified pieces of blocks and generate second image data having the specified number of pixels, and perform the DCT (discrete cosine transform) on the generated second image data as the input image data to produce the DCT coefficients.

4. The image feature generating device according to claim 1, configured so as to partition first image data into specified number of blocks, calculate dominant color information of each of the blocks obtained by the partitioning, and generate second image data having a specified size using the dominant color information of each of the blocks as a pixel, and perform the DCT (discrete cosine transform) on the generated second image data as the input image data to produce the DCT coefficients.

5. The image feature generating device according to claim 2, wherein the second image data has a specified size of 8 pixels×

8 pixels.

6. The image feature generating device according to claim 3, wherein the first image data is partitioned into 8×

8 blocks to generate the second image data having a specified size of 8 pixels×

8 pixels.

7. The image feature generating device according to claim 4, wherein the first image data is partitioned into 8×

8 blocks to generate the second image data having a specified size of 8 pixels×

8 pixels using the dominant color information of each of the blocks obtained by the partitioning.

8. An image retrieving for retrieving image data, using the image feature descriptors generated by the image feature generating device according to claim 1.

9. An image retrieving device configured so as to store in advance image feature descriptors for reference, generate image feature descriptors for inquiry from image data input for inquiry, and calculate similarity between the image feature descriptors for inquiry and the image feature descriptors for reference, wherein the image feature descriptors for inquiry and the image feature descriptors for reference each are generated by the image feature generating device according to claim 1.

10. An image retrieving device for retrieving image data being similar to image data for inquiry from image data for reference comprising:

a first coefficient transforming means for performing a DCT (discrete cosine transform) on the image data for reference to produce DCT coefficients thereon, and performing coefficient transform on the produced DCT coefficients in a manner that fine quantization is carried out on a range of a portion of each of the produced DCT coefficients having a small amplitude and coarse quantization is carried out on a range of a portion of each of the produced DCT coefficients having a large amplitude, in order to generate image feature descriptors for reference, using the transformed and quantized DCT coefficients as the image feature descriptors for reference extracted from the image data for reference;

an image feature descriptor storing means for storing the generated image feature descriptor for reference; and

a similarity calculating means for calculating similarity between the input image data for inquiry and the image data for reference, using an image feature descriptors for inquiry and the image feature descriptors for reference stored in said image feature descriptor storing means.

11. The image retrieving device according to claim 10, further comprising:

a second coefficient transforming means for performing a DCT (discrete cosine transform) on the input image data for inquiry to produce DCT coefficients thereon, and performing coefficient transform on the produced DCT coefficients in a manner that fine quantization is carried out on a range of a portion of each of the produced DCT coefficients having a small amplitude and coarse quantization is carried out on a range of a portion of each of the produced DCT coefficients having a large amplitude, in order to generate the image feature descriptors for inquiry, using the transformed and quantized DCT coefficients as the image feature descriptors for inquiry extracted from the input image data for inquiry, and

wherein the image data for reference is stored in an image database.

12. An image feature generating method for generating image feature descriptors for similarity calculation from input image data, the method comprising:

performing a DCT (discrete cosine transform) on the input image data to produce DCT coefficients, and

performing coefficient transform on the produced DCT coefficients in a manner that fine quantization is carried out on a range of a portion of each of the produced DCT coefficients having a small amplitude and coarse quantization is carried out on a range of a portion of each of the produced DCT coefficients having a large amplitude, in order to generate the image feature descriptors, using the transformed and quantized DCT coefficients as the image feature descriptors.

15. A storage medium having stored thereon computer-executable instructions for causing a computer to execute the image feature generating method according to claim 12.

13. An image retrieving method for retrieving image data being similar to image data for inquiry from image data for reference, the method comprising:

performing a DCT (discrete cosine transform) on the image data for reference to produce DCT coefficients thereon, and performing coefficient transform on the produced DCT coefficients in a manner that fine quantization is carried out on a range of a portion of each of the produced DCT coefficients having a small amplitude and coarse quantization is carried out on a range of a portion of each of the produced DCT coefficients having a large amplitude, in order to generate image feature descriptors for reference, using the transformed and quantized DCT coefficients as the image feature descriptors for reference extracted from the image data for reference;

storing the generated image feature descriptor for reference; and

calculating similarity between the input image data for inquiry and the image data for reference, using an image feature descriptors for inquiry and the stored image feature descriptors for reference.

14. The image retrieving method according to claim 13, further comprising:

performing a DCT (discrete cosine transform) on the input image data for inquiry to produce DCT coefficients thereon, and performing coefficient transform on the produced DCT coefficients in a manner that fine quantization is carried out on a range of a portion of each of the produced DCT coefficients having a small amplitude and coarse quantization is carried out on a range of a portion of each of the produced DCT coefficients having a large amplitude, in order to generate the image feature descriptors for inquiry, using the transformed and quantized DCT coefficients as the image feature descriptors for inquiry extracted from the input image data for inquiry, and

wherein the image data for reference is stored in an image database.

16. A storage medium having stored thereon computer-executable instructions for causing a computer to execute the image retrieving method according to claim 13.

17. The storage medium according to claim 16, further including computer-executable instructions for causing the computer to execute the following image retrieving steps:

performing a DCT (discrete cosine transform) on the input image data for inquiry to produce DCT coefficients thereon, and performing coefficient transform on the produced DCT coefficients in a manner that fine quantization is carried out on a range of a portion of each of the produced DCT coefficients having a small amplitude and coarse quantization is carried out on a range of a portion of each of the produced DCT coefficients having a large amplitude, in order to generate the image feature descriptors for inquiry, using the transformed and quantized DCT coefficients as the image feature descriptors for inquiry extracted from the input image data for inquiry, and

wherein the image data for reference is stored in an image database.