Probability models for highly parallel image coding architecture

A key aspect of image coding systems is the probability model employed to code data. The more precise estimates inferred by model, higher efficiency achieved. In general, models adjust after every new symbol. main difficulty apply such a strategy highly parallel engine that many symbols are coded simultaneously, so adaptation requires different approach. in previous works utilizes stationary collected priori from training set. Its drawback statistics dependent type, images require sets. This work introduces two for architecture that, similarly conventional systems, adapt probabilities while One proposed through finite state machine, other employs already via sliding window. Experimental results indicate latter approach improves performance achieved models, including JPEG2000 and High Throughput JPEG2000, at medium high rates with only slight increase computational complexity.