http://www.ece.stevens-tech.edu/~msync
Speaker: Xiaolin Wu
Time: October
24, 2001, 3:00pm – 4:00 pm
Venue: Burchard
213
Abstract:
A key to the success of any signal compression algorithm
is its ability to separate transient behavior from long term trends. For image/video compression this means the
separation of edge structures from background.
Wavelet transform can represent an image in good time resolution at higher
frequency, facilitating compact description of edges. In this talk, I will examine two approaches of coding edges in
time-frequency domain. The first
approach is high-order context modeling and adaptive entropy coding of wavelet
coefficients. The second approach is explicit multiscale edge modeling in
conjunction with wavelet transform. The
first approach has led to the best compression results in the literature. The research on the second approach is still
at an early stage, but it has opened up some interesting new applications
besides compression, such as content-based retrieval and image analysis
directly on compressed data.
Dr. Xiaolin Wu is a professor at the Department of
Computer and Information Science, Polytechnic University Brooklyn, NY 11201
Host: Prof. K.P. Subbalakshmi