Skip to Main Content
This paper presents a t+2D analysis-synthesis (A-S) video coder which takes into account both motion and geometry. The principle is to pre-distort a group of frames to align the temporal and spatial correlations along the three axis. Motion and geometry are estimated through similar procedures. Distorted frames are obtained through the use of a single operator combining both motion and geometry compensation. The compensated volume is then encoded by a simple 3D separable wavelet coder. The receiver finally decodes the volume and synthesizes original frames by inverting the compensation operators. Using active meshes to represent both the motion and the geometry, compression results point out a significant visual improvement of reconstructed contours compared to an Hybrid coder and a previous A-S coder with no geometry estimation.