OPTIMAL PREDICTIVE DESIGN OF BOOLEAN AND ORDER STATISTICS BASED FILTERS

摘要

This paper first reviews several techniques for optimal predictive design of Boolean and L-filters, which proved successful in lossless audio and image compression. We show that two well known nonlinear predictors, namely the Median Adaptive Predictor and the Gradient Adaptive Predictor, which are used in state of the art lossless compression are particular cases of our FSM-L predictors. Including the GAP predictor parameters as initial conditions in our adaptive FSM-L prediction scheme is shown to consistently improve the compression performance, with no increase in the overall complexity.