Abstract
In this paper, a rate-distortion framework is proposed to define a jointly optimal displacement vector field estimation • DVFE • and quadtree segmentation technique for video coding. This technique achieves maximum reconstructed image quality under the constraint of a target bit rate for the coding of the displacement vector field, quadtree segmentation information, and the residual signal. A fast hierarchical motion estimation scheme as well as a face model-assisted conditional optimization strategy is proposed to drastically reduce the large computation cost required in the R-D optimization process. A novel skin-color based face detection scheme is also proposed for fast locating face regions. The simulation results show that the proposed method can achieve more than 0.5 dB PSNR quality improvement over the H.263 TMN5 video codec at an acceptable extra computation cost.