A lossless coding scheme using predictors and arithmetic code optimized iteratively for each image

摘要

We previously proposed a novel lossless image coding scheme using minimum-rate predictors and variable length codes (VLCs). In this scheme, linear predictors are designed for each image so that a total coding rate estimated on Gaussian models can have a minimum. Therefore coding performance deteriorates when probability density functions (PDFs) of actual prediction errors do not conform to the models. To cope with this problem, in this paper, we introduce new PDF models based on generalized Gaussian functions and iteratively optimize not only predictors but also parameters of the models. Moreover, adaptive arithmetic code is employed in stead of the VLCs to improve coding efficiency in fiat areas of images. Simulation results indicate that the proposed scheme achieves higher coding performance than the state-of-the-art lossless coding schemes for most natural images.