Three-component image compression with frequency-weighted texture coding and edge modeling

Coding methods are presented for the three-component image analysis-synthesis system. The three components contain low-pass, edge, and texture information. The edge component is coded using a parametrized model, while the low-pass and texture components are jointly coded using subband coding with frequency-weighted bit allocation. Coding simulation results show how misleading objective measures can be, particularly regarding edge distortions such as blurriness and ringing. By maintaining edge fidelity, the thee-component coder outperforms the JPEG (Joint Photographic Experts Group) and subband coders at very low bit rates.<<ETX>>