This thesis presents a shape-adaptive vertex-based binary shape coding scheme which achieves a high coding efficiency.
First, a new vertex selection method is proposed for reducing reconstruction errors and bit rates. Three new techniques - vertex adjustment, one-dimensional vertex selection and vertex reordering - are applied after the conventional iterated refinement method. The vertex adjustment method reduces the reconstruction error significantly, while the one-dimensional vertex selection and the vertex reordering method reduce bit-rates when encoding the vertex position.
Second, efficient vertex position encoding methods are proposed. A cumulative segment-based coding method is proposed for encoding initial vertices. And coding method for one-dimensional vertex is also proposed. Furthermore, a conditional differential chain coding is presented for efficient lossless coding of contours. It uses the conditional probability of directional symbols depending on the context of preceding contour directions.
Lastly, the shape-adaptive vertex-based binary shape coding scheme including all techniques proposed in this thesis is presented. The vertex adjustment reduces the reconstruction errors in various contour types, whereas the performances of other proposed techniques for efficient vertex encoding depend on the characteristics of contour.
Experimental results show that the shape-adaptive vertex-based coding scheme reduces coded bits not only by 15~30% compared with the conventional vertex-based coding composed of the iterated refinement method and the object-adaptive vertex encoding, but also up to 15% compared with the context-based arithmetic encoding(CAE) of MPEG-4 Video VM(Verification Model) in intra-coded picture coding.
