Improved motion estimation by spatial transformations and overlap

We propose a new motion model that applies overlapped motion compensation to a motion field generated by spatial transformations based motion estimation. This overlapped spatial transformations (OST) model successfully addresses the inability of the conventional block-matching technique to compensate for complex motion and inside-block motion field discontinuities. For the best performance the overlap technique must be used in both the motion estimation and compensation processes. Simulation results show that, for the tested image sequences, the quality of the motion compensated predicted images, measured in PSNR to the original images, on the average, is increased by 1.5 dB over overlapped and non-overlapped block-matching motion estimation, for the same generated motion vectors overhead. Combining low prediction error and smooth predicted images this model is suitable for very low bit rate video coding.