THE DECISION ORIENTED APPROACH TO INFORMATION-PRESERVING DATA COMPRESSION OF DIGITIZED IMAGES

An information preserving data compression technique using two-dimensional DPCM based on the Markov mesh model that employs maximum likelihood prediction of the image pixel and encodes the differences via Gordon-Block-Bit-Plane (GBBP) encoding is developed. This technique is found to be very efficient in that it requires a bit rate slightly greater than the entropy of the source.

For images with a large number of quantization levels, maximum likelihood prediction can be cumbersome to implement. For this reason, a suboptimal procedure called Differential Bit Plane Coding is investigated. This procedure is easily implemented, even for a large number of quantization levels, and is reasonably efficient.

An information reducing technique based on modifying Differential Bit Plane Coding is studied. This technique is partially information preserving and tends to preserve important features such as object edges.