Key-frame radial projection for robust video hashing
C. De Roover, Christophe De Vleeschouwer, Frédéric Lefèbvre, Benoit M. M. Macq · DIAL (Catholic University of Leuven) · 2005
Robust signal hashing defines a feature vector that characterizes the signal, independently of non-significant distortions of its content, due for example to compression. Our paper first proposes a robust image hashing algorithm, based on radial projection of the image pixels, and evaluates its robustness and discriminating capabilities. The method appears to generate similar feature vectors for visually equivalent images, while resulting in distinct vectors for two different images. Robust image hashing is then extended to video sequences by selecting key-frames, and by extracting a feature vector for each of these frames. Experiments demonstrate that the proposed approach is able to characterize the video sequence content, while providing good robustness towards spatial or temporal distortions.