Block-oriented coding methods transmit the motion and colour information of each block, while object-oriented coding method, additionally, has to transmit shape information of each object.
In this thesis, a reduction method of the shape information in object-oriented sequence image coding is proposed. For the reduction, the proposed method utilizes both the motion information describing the whole object movement and the characteristics of the error regions occurred by the motion compensated prediction of object shape. It is composed of a thresholding operation and a contour coding of errorneous regions. In the thresholding operation, the prediction error regions insensitive to human eyes are eliminated. For the contour coding, three points including the maximum error point between present and predicted contour are used for contour approximation instead of polygon-spline approximation.
Simulation results show that the proposed method trasmits the shape information with less data than the one in the conventional scheme.