Adaptive predictive image coding using two-dimensional multiplicative autoregressive models

Studies adaptive predictive coding of digitized images using novel two-dimensional multiplicative autoregressive (MAR) models. A general stability theorem for 2-D MAR image models with causal nonsymmetric half-plane (NSHP) supports is presented, and two different models are introduced. A major advantage of using 2-D MAR models, as opposed to general 2-D models, is the ability to guarantee predictor stability for NSHP support regions. The performance of 2-D MAR predictive coders is tested on five different images. The results of these studies demonstrate that, for coding of digitized images, 2-D MAR model-based predictive coders offer an effective alternative to currently available techniques.<>