Digital watermarking of MPEG-4 facial animation parameters

Abstract

The enforcement of intellectual property rights is difficult for digital data. One possibility of supporting enforcement is the embedding of digital watermarks containing information about copyright owner and/or receiver of the data. Watermarking methods have already been presented for audio, images, video, and polygonal 3D models. In this paper, we present a method for digital watermarking of MPEG-4 facial animation parameter data sets. The watermarks are additively embedded into the parameter values and can be retrieved either from the watermarked parameters or from video sequences rendered using the watermarked animation parameters for head animation. For retrieval from rendered sequences, the facial animation parameters have to be estimated first. We use a model-based approach for the estimation of the facial parameters that combines a motion model of an explicit 3D textured wireframe with the optical flow constraint from the video data. This leads to a linear algorithm that is robustly solved in a hierarchical framework with low computational complexity. Experimental results confirm the applicability of the presented watermarking technique.