Image Retrieval Based on RST Invariant Features Extracted from Scale Invariant Keypoints

In content-based image retrieval systems, the invariance to geometrical transformations is one of the most desired properties. In this paper, a kind of rotation, scaling and translation (RST) invariant feature for image retrieval is investigated, and a new method is proposed to extract this type of feature. The proposed scheme first detects the scale invariant key points in images, and then utilizes the translation and rotation invariance properties of Burkhardt’s features to construct a local rotation and translation (RT) descriptor. Finally, the RST invariant features are extracted from the key points based on the local descriptor. Moreover, we take the structural information into account and combine it with the histogram descriptor. By combining these techniques, we can effectively retrieve both the RST transformed images and the similar images of the query image. Experimental results demonstrate the effectiveness of the proposed scheme by comparing it with other methods.