A geometric data structure applicable to image mining and retrieval
ICIAR'10 Proceedings of the 7th international conference on Image Analysis and Recognition - Volume Part I
| Hi-index | 0.00 |
Interest points are local properties of an image with high informational content. In this paper, we present a new method for image retrieval based on color and distribution of interest points. Firstly, we use Harris corner detector to extract interest points. Secondly, we design a 2D histogram based on interest points to describe an image, which not only takes the local color feature into consideration, but also the space distribution information of interest points. With robustness to image rotation and translation, the 2D histogram avoids the shortcoming of losing the space distribution information in color histogram. Finally,we employ the weighted distance between two 2D histograms to discriminate the similarity between twoimages. Lots of experimental results based on a database containing 1100 images demonstrate our proposed method has good performance.