Segmented image coding: Techniques and experimental results

摘要

This paper experimentally compares several variants of segmented image coding (SIC), some of which are new. First it compares two methods for approximating the image intensity, with respect to quality of the reconstructed image, computational complexity and memory requirements. An extension of an existing segmentation algorithm (edgmentation) is introduced and compared with two other segmentation algorithms suitable for SIC. Finally, a hybrid scheme is proposed which utilizes both SIC and block transform coding (BTC) methods. The hybrid scheme has 35% less computational complexity than the full SIC approach, while it retains all the edge information produced by the segmentation algorithm.