Perceptual video hashing using 3D-radial projection technique

Abstract

A function that outputs a feature vector from the perceptual contents of the input video is called a perceptual video hashing function and the output feature vector that characterizes the perceptual contents of the input video is called the perceptual video hash. This hash must be robust to the manipulations that preserves the perceptual contents of the video and fragile to the modifications that vary the perceptual contents of the video. The wide scale applications of perceptual hash in the world of multimedia, such as video authentication, video copyright protection, and video retrieval, emphasize the importance of this area of research. This work generates a perceptual hash from the video using the 3D-radial projection of the pixels and assesses the differentiating capabilities and the perceptual robustness of the hash generated. This method is the 3D extension of the 2D radial projection based image hashing. In this work, the variance of the luminance values of the projected pixels is calculated for each randomly generated sub-tensor. The variance of all the sub-tensors are averaged along the second dimension and projected onto the discrete cosine transform (DCT) basis. The performance measure of the proposed work is assessed using the receiver operating characteristic (ROC) curves. Simulation results indicate that the performance of the proposed method is satisfactory for both the content-preserving and the content changing attacks.