Compression of image data

摘要

Image data such as binary mask image data is compressed by partitioning the data into blocks. An arithmetic encoder is initiated for each block and terminated at the end of the block. For each block, there is a fixed template/PDF table pair model which is chosen at the beginning of the block according to motion compensation and other inputs. One model has a template with spatially-correlating samples only, the other having both temporally and spatially-correlating samples. Estimations are made for values of samples which border the current block according to simple rules for the particular template.