Robust Image Copy Detection Using Local Invariant Feature

This paper proposes a novel geometric distortion resilient image copy detection scheme based on Scale Invariant Feature Transform (SIFT) detector. By using the SIFT detector, the proposed copy detection scheme first construct a series of robust, homogenous, and larger size circular patches. And then, the cirque track division strategy and ordinal measure concept are introduced to generate a cirque-based ordinal measure feature vector for each circular patch. Besides, the ROC graph and MAP probability are utilized to estimate the two parameters (vector dimension and detection threshold) respectively. Experimental results and the related analysis show that the proposed scheme is robust to most of geometric and photometric distortions.