SIFT-CCH: Increasing the SIFT distinctness by Color Co-occurrence Histograms

Describing regions in a distinctive way, in order to find correct correspondences in images of two separated views, represents a complex and essential task of computer vision. Until now, SIFT (Scale Invariant Feature Transform) has been proven to be the most reliable descriptor among the others. One of the main drawbacks of SIFT is its vulnerability to color images, being designed mainly for the gray images. To overcome this problem and also to increase the overall distinctness of the SIFT in this paper we introduce a new descriptor that combines the SIFT approach with the color co-occurrence histograms (CCH), a concept used extensively in color texture retrieval and object recognition applications. We evaluate the new descriptor in the context of image matching. The experimental results show that our descriptor outperforms the original version, detecting an important number of additional correct matched feature points while the mismatch ratio remains constant.