Region coding and detail compensation

Abstract

An approach is proposed for segmented image coding which avoids the redundant and false-edges that appear in the late stages of hierarchical segmentation. The recursive shortest spanning tree (RSST) algorithm is used to obtain regions describing the main objects. Significant small features are then extracted from the original image to compensate for the coarse segmented image. The method of extracting details is based on mathematical morphology and is a modification of a detail extracting method described in a previous study. This approach produced better visual image quality than using the segmentation-only algorithms especially in the case of low bit-rate coding.