Adaptive contour representation based on morphological segmentation

Abstract

The contour coding in the segmentation-based coding usually occupies the biggest part in the encoded bitstream. In this paper, contour segments are adaptively encoded by considering neighbor regions. Encoded bitrate is reduced without loss of the subjective image quality. After morphological segmentation of the image using the watershed algorithm, the contour segments are classified according to the contrast of the adjacent regions. Then, the contour segments with low contrast are highly compressed with various simplification such as polygonal approximation and morphological low pass filtering. Experimental results show outstanding performance for the very low bitrate video coding.