Applying binary partitioning to weighted finite automata for image compression

Abstract

Fractal-based image compression techniques give efficient decoding time with primitive hardware requirements, which favors real-time communication purposes. One such technique, the Weighted Finite Automata (WFA) is studied on grayscale images. An improved image partitioning technique - the binary or bin-tree partitioning - is tested on the WFA encoding method. Experimental results show that binary partitioning consistently gives higher compression ratios than the conventional quad-tree partitioning method. Moreover, the ability to decode images progressively rendering finer and finer details can be used to display the image over a congested and loss-prone network such as the Image Transport Protocol (ITP) for the Internet, as well as to pave way for multi-layered error protection over an often unreliable networking environment such as the UDP. ©2004 IEEE.