Robust video hashing using the 3D-DWT

Robust or perceptual hash functions for video extract hashes on the basis of the perceptual contents of videos. Besides being sensitive to content differences in distinct videos, such a hash function should be robust against the content-preserving operations. This paper presents a solution for the hashing of videos at the group-of-frames level using the 3-dimensional discrete wavelet transform. A hash of a group-of-frames is computed from the spatio-temporal low-pass band of the transform. The local variances of the wavelet coefficients in the band are used to derive the hash. Experimental results demonstrate the robustness of the hash function against content-preserving operations and the sensitivity to content differences.