Copy image detection based on local keypoints

Detecting copy images of a query image in large scale image collections is a very important task for many applications, such as copyright violations detection and copy image filtering in the results of image retrieval. In this paper, a novel method is proposed in which each image is represented as a set of local keypoints. The local keypoint is characterized by a compact fingerprint to minimize the effect of color changing. This keypoint descriptor is more compact than the feature vector descriptor. Hamming distance is used to measure the similarity of two fingerprints. To retrieve a fingerprint quickly in one large scale fingerprint collection, a fast retrieval method is adapted to construct sorted tables of the fingerprints by grouping the bits of fingerprint. This retrieval method has low complexity. The experimental results validate the effectiveness of the proposed algorithms.