A fuzzy gradient-adaptive lossy predictive coding technique

In this paper a new adaptive fuzzy predictive coding system is introduced. The proposed coder employs the adaptive fuzzy prediction methodology developed in [Tian-Hu Yu, 1998]. This results in better prediction of smooth as well as edge regions. In addition, the proposed coder adopts a novel fuzzy gradient-adaptive quantization scheme that switches between three well-designed nonuniform quantizers depending on the local gradient of the pixel to be coded. This, in turn, leads to reduced quantization errors in both smooth and edge regions and consequently higher perceptual quality of reconstructed images is achieved.