Real-time estimation of long-term 3-D motion parameters for SNHC face animation and model-based coding applications

摘要: We present two recursive methods for the real-time estimation of long-term three-dimensional (3-D) motion parameters from monocular image sequences suitable synthetic/natural hybrid coding face animation and model-based applications. Based on feature point extractions in energy frame, 3-D a human are estimated with predictive approach. The first method uses linear least squares approach second employs nonlinear extended Kalman filter, which does not rely linearized model motion. Both perform prediction correction loop at every time step. Compared to other described literature, structure proposed process solves problem error accumulation estimation. This makes stable consistent over long periods. Experimental results presented synthetic data real sequences, demonstrate performance compare approaches.