SEECS Colloquium

Content-based access of image and video data is an interdisciplinary area involving research in information retrieval, multimedia processing, computer vision, and machine learning, etc. I will present my recent work in content-based image retrieval (CBIR) and content-sensitive video streaming.

To bridge the semantic gap, we put the user in the loop during the retrieval process. To deal with positive and negative user feedbacks in a principled way, we model a relevance feedback process as a small sample asymmetric learning problem, and propose the BiasMap algorithm as a general solution. Nonlinear capability is achieved by using either a kernel trick or a boosting approach. As a discriminative density estimator, BiasMap can be applied to other small sample classification problems. For global image representation, I proposed the Water-Filling algorithm for extracting edge-based structural features to complement existing texture and shape features; For local modeling, I experimented with a factorized probabilistic local appearance model for object detection in cluttered scenes. Inspired by pseudoclassification concept from traditional information retrieval research, I proposed a scheme for interaction and unification of keywords and low-level features for databases with limited textual annotations. My work in content-sensitive video streaming resulted in two modeling schemes and optimization algorithms for content-sensitive video streaming under channel and buffer constraints. Detailed information and preprints are available at http://www.ifp.uiuc.edu/~xzhou2

Xiang Sean Zhou received his Bachelor's degrees in Automatic Control and Economics and Management in 1993 from Tsinghua University, China; where he also studied Economics for two years in a PhD program before he joined the University of Cincinnati in 1996 and received his M.S. degree in Electrical and Computer Engineering in 1998. Currently he is with the Beckman Institute for Advanced Science and Technology at University of Illinois at Urbana Champaign (UIUC), where he is a research assistant and a graduating PhD student. His research interests include computer vision, pattern recognition, machine learning, signal and image processing, and multimedia information retrieval. He has published papers in Optical Engineering, IEEE Multimedia, Pattern Recognition Letters, IEEE Transactions on Circuits and Systems for Video Technology, and various conference proceedings. He was the recipient of numerous scholarships and awards from Tsinghua University. In 2001, he was awarded the M. E. Van Valkenburg Fellowship Award, an award given to one or two PhD students in the ECE department of UIUC each year "for demonstrated excellence in research in the areas of circuits, systems, or computers."