Simplification of image partitions for region-based image coding

In this paper we address the problem of lossy image partition encoding, and compare the compression efficiency of two contour coding techniques associated with appropriate contour simplification methods: the majority filter for the case of chain coding; the reduction of the number of transition points for the encoding technique based on transition points. The experimental results show that chain coding only perform better when unconstrained simplification is allowed, i.e., when the reconstruction error is unlimited. In all other cases, for which the error is bounded, the encoding technique based on transition points shows superior compression efficiency.