Traffic video data protection based on a simultaneous autoregressive image model

In this work, we propose a hybrid domain approach for traffic video data protection. More specifically, using a hierarchical Bayesian model, we propose a new watermarking scheme for copyright protection as well as proof of ownership for video traffic data. During embedding, the hidden information is added in the wavelet domain exploiting the properties of robustness and imperceptibility of DWT (Discrete Wavelet Transform). Subsequently, we propose to model every frame in spatial domain with a simultaneous autoregressive (SAR) process, which is driven by a residual term described by known heavy tailed distributions. Based on these statistical image models we derive optimal detectors for the problem at hand. Numerical experiments and conclusions are given to show the performance of our scheme and the comparison between the proposed statistical models.