Apparatus and method for retrieving image

1. An image retrieval method of a computer for retrieving similar objects, comprising:

reading from a first database a plurality of images of an image group mapped to a multidimensional object as a retrieval target, and extracting an image feature quantity of the read plurality of images, the first database storing the plurality of images depicting the multidimensional object as the retrieval target from a plurality of orthoprojection viewpoints of the multidimensional object in different directions;

reading from a second database a plurality of images of an image group mapped to a multidimensional object as a search key, and extracting an image feature quantity of the read plurality of images, the second database storing the plurality of images depicting the multi-dimensional object as the search key from a plurality of orthoprojection viewpoints in different directions;

determining a similarity between the image group mapped to the multidimensional object as the search key and the image group mapped to the multidimensional object as the retrieval target, based on the image feature quantity of the plurality of images of the image group mapped to the multidimensional object as the retrieval target and the image feature quantity of the plurality of images of the image group mapped to the multidimensional object as the search key; and

displaying, on an output device, as retrieval results the plurality of images of the image group mapped to the multidimensional object as the retrieval target in an order of similarity of the multidimensional object as the retrieval target based on the determined similarity,

wherein the retrieval results are presented by displaying the multidimensional object as the search key in a cubic diagram using the plurality of images of the multidimensional object as the search key, and by displaying the multidimensional object as the retrieval target in another cubic diagram using the plurality of images of the multidimensional object as the retrieval target.

2. The image retrieval method according to claim 1 , wherein the retrieval results are presented by determining a correspondence relationship representing a similarity between an image resulting from depicting the multidimensional object as the search key from one orthoprojection viewpoint in one direction and an image resulting from depicting the multidimensional object as the retrieval target from another orthoprojection viewpoint in another direction, based on the image feature quantity of the plurality of images of the image group mapped to the multidimensional object as the retrieval target and the image feature quantity of the plurality of images of the image group mapped to the multidimensional object as the search key, and by rearranging and displaying the plurality of images of the multidimensional object as the retrieval target in accordance with the correspondence relationship so that the plurality of images of the multidimensional object as the retrieval target are similar to the plurality of images of the multidimensional object as the search key.

3. The image retrieval method according to claim 1 , wherein the cubic diagram of the multidimensional object as the retrieval target is configured to be rotated in synchronization with and in the same direction as the cubic diagram of the multidimensional object as the search key.

4. The image retrieval method according to claim 1 , wherein in order to determine the similarity between the image group mapped to the multidimensional object as the search key and the image group mapped to the multidimensional object as the retrieval target, all combinations of each of the images of the image group mapped to the multidimensional object as the search key and each of the images of the image group mapped to the multidimensional object as the retrieval target are calculated in terms of similarity, a correspondence relationship of the images providing the maximum sum of the similarities is determined, and the mean value of the similarities of the images in the correspondence relationship providing the maximum sum of the similarities is set to be a similarity between the multidimensional object as the retrieval target and the multidimensional object as the search key.

5. The image retrieval method according to claim 1 , wherein in order to determine the similarity between the image group mapped to the multidimensional object as the search key and the image group mapped to the multidimensional object as the retrieval target, all combinations of each of the images of the image group mapped to the multidimensional object as the search key and each of the images of the image group mapped to the multidimensional object as the retrieval target are calculated in terms of similarity, a correspondence relationship of the images providing the maximum sum of the similarities is determined, and the maximum similarity of the images in the correspondence relationship providing the maximum sum of the similarities is set to be a similarity between the multidimensional object as the retrieval target and the multidimensional object as the search key.

6. An image retrieval apparatus, comprising:

a memory; and

a processor that executes a procedure in the memory, the procedure including:

a first image feature extracting process configured to read from a first database a plurality of images of an image group mapped to a multidimensional object as a retrieval target, and extracting an image feature quantity of the read plurality of images, the first database storing the plurality of images depicting the multi-dimensional object as the retrieval target from a plurality of orthoprojection viewpoints in different directions;

a second image feature extracting process configured to read from a second database a plurality of images of an image group mapped to a multi-dimensional object as a search key, and extracting an image feature quantity of the read plurality of images, the second database storing the plurality of images depicting the multi-dimensional object as the search key from a plurality of orthoprojection viewpoints in different directions;

a similarity calculating process configured to calculate a similarity between the image group mapped to the multidimensional object as the search key and the image group mapped to the multidimensional object as the retrieval target, based on the image feature quantity of the multidimensional object as the retrieval target and the image feature quantity of the multidimensional object as the search key; and

a retrieval result displaying process configured to display, on an output device, as retrieval results the plurality of images of the image group mapped to the multidimensional object as the retrieval target in the order of similarity of the multidimensional object as the retrieval target based on the similarity,

wherein the retrieval results are presented by displaying the multidimensional object as the search key in a cubic diagram using the plurality of images of the multidimensional object as the search key, and by displaying the multidimensional object as the retrieval target in another cubic diagram using the plurality of images of the multidimensional object as the retrieval target.

7. The image retrieval apparatus according to claim 6 , wherein the retrieval result displaying process determines a correspondence relationship representing a similarity between an image resulting from depicting the multidimensional object as the search key from one orthoprojection viewpoint in one direction and an image resulting from depicting the multidimensional object as the retrieval target from another orthoprojection viewpoint in another direction, based on the image feature quantity of the plurality of images of the image group mapped to the multidimensional object as the retrieval target and the image feature quantity of the plurality of images of the image group mapped to the multidimensional object as the search key, and rearranges and displays the plurality of images of the multidimensional object as the retrieval target in accordance with the correspondence relationship so that the plurality of images of the multidimensional object as the retrieval target are similar to the plurality of images of the multidimensional object as the search key.

8. The image retrieval apparatus according to claim 6 , wherein the retrieval result display process is configured to rotate the cubic diagram of the multidimensional object as the retrieval target in synchronization with and in the same direction as the cubic diagram of the multidimensional object as the search key.

9. The image retrieval apparatus according to claim 6 , wherein the similarity calculating process calculates similarities of all combinations of each of the images of the image group mapped to the multidimensional object as the search key and each of the images of the image group mapped to the multidimensional object as the retrieval target, determines a correspondence relationship of the images providing the maximum sum of the similarities, and sets the mean value of the similarities of the images in the correspondence relationship providing the maximum sum of the similarities to be a similarity between the multidimensional object as the retrieval target and the multidimensional object as the search key.

10. The image retrieval apparatus according to claim 6 , wherein the similarity calculating process calculates similarities of all combinations of each of the images of the image group mapped to the multidimensional object as the search key and each of the images of the image group mapped to the multidimensional object as the retrieval target, determines a correspondence relationship of the images providing the maximum sum of the similarities, and sets the maximum similarity of the images in the correspondence relationship providing the maximum sum of the similarities to be a similarity between the multidimensional object as the retrieval target and the multidimensional object as the search key.

11. A non-transitory computer-readable storage medium that stores a program causing a computer to perform a process of retrieving similar multidimensional objects, the process comprising:

reading from a first database a plurality of images of an image group mapped to a multidimensional object as a retrieval target, and extracting an image feature quantity of the read plurality of images, the first database storing the plurality of images depicting the multi-dimensional object as the retrieval target from a plurality of orthoprojection viewpoints in different directions;

reading from a second database a plurality of images of an image group mapped to a multidimensional object as a search key, and extracting an image feature quantity of the read plurality of images, the second database storing the plurality of images depicting the multi-dimensional object as the search key from a plurality of orthoprojection viewpoints in different directions;

determining a similarity between the image group mapped to the multidimensional object as the search key and the image group mapped to the multidimensional object as the retrieval target, based on the image feature quantity of the plurality of images of the image group mapped to the multidimensional object as the retrieval target and the image feature quantity of the plurality of images of the image group mapped to the multidimensional object as the search key; and

displaying, on an output device, as retrieval results, the plurality of images of the image group mapped to the multidimensional object as the retrieval target in the order of similarity of the multidimensional object as the retrieval target based on the determined similarity,

wherein the retrieval results are presented by displaying the multidimensional object as the search key in a cubic diagram using the plurality of images of the multidimensional object as the search key, and by displaying the multidimensional object as the retrieval target in another cubic diagram using the plurality of images of the multidimensional object as the retrieval target.

12. The non-transitory computer-readable storage medium according to claim 11 , wherein the retrieval results are presented by determining a correspondence relationship representing a similarity between an image resulting from depicting the multidimensional object as the search key from one orthoprojection viewpoint in one direction and an image resulting from depicting the multidimensional object as the retrieval target from another orthoprojection viewpoint in another direction, based on the image feature quantity of the plurality of images of the image group mapped to the multidimensional object as the retrieval target and the image feature quantity of the plurality of images of the image group mapped to the multidimensional object as the search key, and by rearranging and displaying the plurality of images of the multidimensional object as the retrieval target in accordance with the correspondence relationship so that the plurality of images of the multidimensional object as the retrieval target are similar to the plurality of images of the multidimensional object as the search key.

13. The non-transitory computer-readable storage medium according to claim 11 , wherein the cubic diagram of the multidimensional object as the retrieval target is rotated in synchronization with and in the same direction as the cubic diagram of the multidimensional object as the search key.

14. The non-transitory computer-readable storage medium according to claim 11 , wherein in order to determine the similarity between the image group mapped to the multidimensional object as the search key and the image group mapped to the multidimensional object as the retrieval target, all combinations of each of the images of the image group mapped to the multidimensional object as the search key and each of the images of the image group mapped to the multidimensional object as the retrieval target are calculated in terms of similarity, a correspondence relationship of the images providing the maximum sum of the similarities is determined, and the mean value of the similarities of the images in the correspondence relationship providing the maximum sum of the similarities is set to be a similarity between the multidimensional object as the retrieval target and the multidimensional object as the search key.

15. The non-transitory computer-readable storage medium according to claim 11 , wherein in order to determine the similarity between the image group mapped to the multidimensional object as the search key and the image group mapped to the multidimensional object as the retrieval target, all combinations of each of the images of the image group mapped to the multidimensional object as the search key and each of the images of the image group mapped to the multidimensional object as the retrieval target are calculated in terms of similarity, a correspondence relationship of the images providing the maximum sum of the similarities is determined, and the maximum similarity of the images in the correspondence relationship providing the maximum sum of the similarities is set to be a similarity between the multidimensional object as the retrieval target and the multidimensional object as the search key.