Regular Linear Phase Perfect Reconstruction Filter Banks and Application in Image Compression

Dr. Soontorn Oraintara

Professor
Dept. of Electrical Engineering
University of Texas, Arlington

Friday, April 20th, 3:00 PM, ENS 302

oraintar@uta.edu,


Abstract

In recent years, the continued rapid growth of the internet and wireless communications have been creating new and improved services in data access around the world. With the ubiquity of the needs in transporting multidimensional data such as images and video signals, compression technology has become an essential tool that contributes to the success of information exchange over very limited-bandwidth channels. Transform-based coder is one of the most popular systems that are currently employed in many coding standards including JPEG and video MPEG by exploiting the limitations of human visual perception. Its linear transformation and quantization enable decorrelation of spatial redundancy inherent in such signals, by taking into account the non-uniform spectral distribution nature.

The challenge in designing a transform for lossy compression with high perceptual quality is to simultaneously obtain several attractive properties in the transform such as its invertibility, smoothness and symmetry of the basis functions and high coding gain. In this research, we study the theory, structure and design method of regular linear phase perfect reconstruction filter banks which can be viewed as a unified tool for constructing such a transform. Several transforms are optimized for the image coding purposes, and a progressive transmission image coder is used to evaluate their performances. Simulation results show that, while maintaining comparable objective coding performance to the existing works, our transforms are found to improve the perceptual quality of the compressed images where the unpleasant artifacts are significantly reduced.

Biography

Prof. Soontorn Oraintara received his B.S. degree from China in 1985 and his M.S. and Ph.D. degree in Electrical Engineering from the Universities of Wisconsin and Boston, respectively, in 1986 and 2000. He is currently an assistant professor at The Univeristy of Texas at Arlington. His research interests include applying filter banks to image compression.


A list of Telecommunications and Signal Processing Seminars is available at from the ECE department Web pages under "Seminars". The Web address for the Telecommunications and Signal Processing Seminars is http://anchovy.ece.utexas.edu/seminars