CAREER: Research and Development of Advanced Visual Information Systems

This research investigates an integrated approach to the design of advanced visual information systems (VIS). The approach explores the maximum synergy among various important VIS design issues - visual data indexing, compression, manipulation, and storage. An integrated feature map combining multiple image features such as texture, color, and shape is used to detect and index image content, and provide useful methods for content-based image query. Compressed-domain techniques are explored for extracting low-level image features. A modified quad-tree data structure is proposed to be used in feature-based segmentation and heterogeneous feature integration. Based on the prior work, efficient image manipulation algorithms (e.g., image filtering and template matching) are also being developed directly in the compressed domain, in order to seek the minimum implementation complexity. Strategies for storing multi-resolution encoded visual data in real-time hierarchical storage systems are also studied. All research results will be incorporated into an large-scale integrated VIS testbed at the Advanced Imag e Lab of Columbia University. On the educational front, new courses and curriculum in the area of visual signal processing and multimedia technology are being developed. An advanced multimedia testbed is also being developed for supporting research and educational activities.