Adaptive context-based arithmetic coding of arbitrary contour maps

An efficient technique for encoding arbitrary contours is presented. It is based on the concept of transition points of a contour map and relies on a new four-symbol adaptive context-based arithmetic encoder (CAE) that calculates contexts in the (binary) domain of the contour map. The results obtained with this new technique, applied to the compression of image partitions, are substantially better than those previously attained, showing a clear superiority over adaptive CAE-based differential chain coding and also over the MPEG-4 shape coding.