A Robust Near-Duplicate Images Detection Approach with Ordinal Measure

Digital image is easy to be acquired, modified and redistributed either lawfully or unlawfully. A large number of similar images waste storage and computing resources, and sometimes it can lead to legal disputes. A robust algorithm of detecting near-duplicate images is proposed based on the image feature coming from ordinal measure. It can be used to detecting images that are altered to original image in contrast, saturation, blur and other visual appearance of color. And, the hash code is introduced to speed up the comparison and retrieval of near-duplicate images. The experiment on the test image sets proves the effectiveness and efficiency of the proposed method.