Defining embedding distortion for motion vector-based video steganography

Abstract

This paper presents an effective methodology for motion vector-based video steganography. The main principle is to design a suitable distortion function expressing the embedding impact on motion vectors by exploiting the spatial-temporal correlation based on the framework of minimal-distortion steganography. Two factors are considered in the proposed distortion function, which are the statistical distribution change (SDC) of motion vectors in spatial-temporal domain and the prediction error change (PEC) caused by modifying the motion vectors. The practical embedding algorithm is implemented using syndrome-trellis codes (STCs). Experimental results show that the proposed method can enhance the security performance significantly compared with other existing motion vector-based video steganographic approaches, while obtaining the higher video coding quality as well.