In recent years, object/region-based video coding schemes have attracted considerable attention in the image coding community. These schemes are expected to yield better coding results, especially at low bit-rates. In this paper a new object/region-based video coding scheme is presented. It is based on a layered object representation for image groups along with affine motion fields. Thus, groups of images are decomposed into a number of object sequences. Each object sequence is motion compensated, transformed and coded separately using a 3-D wavelet transform. The presented results prove that object/region-based coding schemes can outperform the block-based standards in objective (PSNR) and subjective (visual) quality.
展开▼
