VCG uses a state of the art  compression algorithm developed by me, its actually 2-in-1 algorithm and doesboth compression and encoding simultaneously in the same phase. Compressionin the sense, i mean  the transformation of the data in to some form which can be encoded  with lesser number of bits than theoriginal and encoder is just packing of  the transformed data.

            For example, popular techniqueslike LZ, LZW & variants  just transform the data by reducing theredundancy in it, but to achive maximum compression you need some encoderlike arithmatic coding or huffman coding to pack it effectively.

            On the contrary, compression techniques like PPM , DMC &various  models use arithmatic coding techniques as a partof their algorithm  and they don't need a seperate encoding phase. LZ and  its  variants  are almost  twenty years old and they achieve  lesser compression  than PPM & contextbased models, which  are more recent and  proved to be the best.

            Unlike these techniques my algorithmfunctions as an encoder  for  evenly  distributed  data and achieves more compression  for unevenly distributed data by reducing the redundancy in one portion of the data compared to other portions. Same can be achieved by a context based model  but for images  hardly there won't be any real repetitions when compared to text files and they perform poor for images compared to otherfiles.

            VCG's compression algorithmis optimised for images and it achieves best  compression for all types of images. Since i haven't  finished working on this  algorithm  i can't  distribute its source code now, but the source code of other modules of this software are available for research  or educational  purpose for a minimal fee.