Perceptual video hashing based on temporal wavelet transform and random projections with application to indexing and retrieval of near-identical videos

Abstract

A perceptual video hash function extracts a short fixed-length bit string called a perceptual hash on the basis of the visual contents of the video. Such a function should be robust to the content-preserving operations and at the same time, sensitive to the content differences. In this work, the discrete wavelet transform (DWT) along the temporal direction, referred to as the temporal wavelet transform (TWT), is used for generating the temporally informative representative images (TIRIs). The resultant low pass data are projected onto the Achlioptas’s random basis to generate the hash. The TWT and the random projection technique not only reduce the dimensions but also retains the important features. Simulation results show that the proposed algorithm performs better for both the content-preserving and the content changing attacks when compared to that of the existing video hashing algorithms with the added advantage of computational efficiency. The proposed algorithm is applied to the indexing and retrieval of near-identical video application and the performance is evaluated using average precision-recall curves.