A Theoretical Framework for Watermarking of Compressed Signals

Abstract

Nowadays, digital video, audio and multimedia signals are mostly transported, stored, exchanged, broadcast in compressed form. Thus, most of the multimedia processing operations, amongst them watermarking, shall be done in the compressed domain. When the bitrate is small, the quantization steps are coarse, and the inserted watermark may be heavily distorted or visible; moreover, watermark insertion may increase the stream rate, which is most of the time not acceptable. We propose to model the watermark insertion as an optimization problem on a discrete set, with visibility and rate constraints. We give an iterative method to get the optimal watermarked signal. We show by experimental results that this approach increases the robustness of the watermark.