Efficient Techniques for Feature-Based Image/Video Access and Manipulation

Abstract

This paper describes our recent work on content-based visual query with emphasis on automatic visual feature extraction, efficient feature indexing, and feature-assisted search and browsing. Efficient techniques for manipulating compressed videos are included as well. We will discuss several research prototypes, including VisualSeek, which is a Java-based WWW application supporting localized color and spatial similarity retrieval; CVEPS (Compressed Video Editing and Parsing System), which supports feature-based indexing and editing of video in the compressed domain; and a hierarchical news video browsing and indexing system.