Method and apparatus for adaptive facsimile compression using a two

Abstract

Data compression, for either a storage or transmission, of facsimile information is effected employing a two dimensional, non-contiguous prediction matrix. A data stream representing the data to be compressed is input to a buffer which is of a size to store sufficient information for prediction purposes. A predictor is responsive to the buffer for producing a predicted data representation from a plurality of data units comprising a two dimensional matrix. A selector is responsive to the data unit employed in the prediction process for making a select/non-select determination. For those data units which are selected, a comparator compares the predicted status of the data unit with the actual status of the data unit. At least one run length encoder is responsive to the comparator for run length encoding successive correct predictions and a following incorrect prediction. An output buffer is provided for storing the run length encoded output of the run length encoder as well as representations of the non-selected data units. By using plural run length encoders, each can be optimized for the encoded data by correlating prediction difficulty with code length, i.e., easy predictions are encoded by long code words and hard predictions are encoded with short words. Further, the unselected class data units correspond to most difficult predictions and these are not coded.