Content-Based Image Retrieval by Interest Points Matching and Geometric Hashing

This paper presents a content-based image retrieval technique based on interest points matching and geometric hashing. We estimate points with significant luminance variations as interest points. A small region around the interest point is located as an image patch. Low-level features are extracted to describe each image patch. To provide geometric invariant image matching, we index the image patches into a 2-D hash table by geometric hashing technique. Thus, the matching is invariant to global and local geometric transforms. In addition, since we use the image patch to capture the local information, the indexing can effectively handle partial matching. We formulate a matching criterion by weighted voting technique to incorporate the spatial interrelationship into consideration. We have performed a series of experiments to confirm the effectiveness of our method. Images are globally transformed and locally manipulated to examine the efficiency of our indexing scheme. Experimental results indicate satisfactory retrieval in the case of partial matching and geometric transformation.