In a conventional block matching method, prediction efficiency decreases and moreover it becomes difficult to produce a natural predicted image when a block includes an outline of a moving object. It is thought that this problem can be overcome by dividing an image into variable-shape-blocks suitable for its contents From this point of view, this paper proposes a new motion compensation method which partitions the previous image into variable-shape-blocks composed of quadrilateral s so that their boundaries can run parallel with its principal contours. Then texture-mapping using pseudo Affine transformation is carried out in every block to produce a predicted image. Simulation results indicate that this method can provide not only higher prediction efficiency but also more natural predicted images than the conventional block matching method.
抄録全体を表示