Real-Time Image COMDression with Edge Feature for Airborne Reconnaissance

We propose to develop an advanced motion-compensated video compression that uses optical-flow motion estimation and segmentation to high compression rates while maintaining critical image features such as edges and textures. Optical-flow is the apparent two-dimensional flow of intensity changes in a sequence of images. Based upon our earlier work on optical-flow image analysis, an advanced optical-flow segmentation algorithm, using discontinuity information in both velocity and intensity domains, would permit the segmenting of video sequences along naturally occurring motion boundaries (edges). By encoding the boundary information and the velocities of uniformly moving patches in the image, a compression system would make better use of the frame- to-frame coherency in video sequences than do current block-based estimation algorithms. Such an encoding system would result in greatly reduced coding size for video, and would, additionally, preserve edge information in the decompressed video. The inherent parallel nature of the optical-flow algorithm allows for implementation on low-cost analog ASIO, providing an inexpensive, real-time compression engine. During Phase I we will develop a video compression CODEC that uses optical-flow image segmentation, and evaluate its ability to preserve critical image features. In Phase II we will develop a real-time implementation of the CODEC in VLSI hardware.