Conditional variable-length coding for gray-level pictures

This paper presents a new variable-length coding algorithm for bit-rate reduction of digital images. The coding scheme is based on the fact that by taking into account the local conditional statistics of the picture, the coding efficiency can be improved considerably. Results of computer simulations are presented in terms of picture quality and required bit rate. Simulations show that for various Differential Pulse Code Modulation (DPCM) coded pictures, coding efficiency is such that a bit rate below the entropy of the DPCM quantizer output is attained. For a 3-bit DPCM encoded picture, a bit rate of 1.64 bits/picture element is achieved.