A near-duplicate 3D video detection algorithm by using hypercomplex representations

Copyright protection is still a crucial open issue in the 3D video industry with the development of 3DTV coupled with the increasing 3D video spread over the internet. In this paper a novel video fingerprinting is proposed for near-duplicate 3D video detection. Instead of generating fingerprints from the color frames and the depth maps separately, the proposed algorithm processes them in a holistic manner. The hypercomplex representation developed from the RGB and depth components is used to represent the 3D contents as quaternion frames and a novel quaternion centroid of the spatio-temporal gradient orientations is exploited to generate 3D video fingerprints based on these quaternion frames. Comprehensive experiments are conducted to evaluate the performance of the proposed method, and the results show that the proposed near-duplicate 3D video detection algorithm outperforms the state-of-the-art approaches in terms of robustness and discrimination. And the proposed fingerprints are also more compact than the existing approaches.