The objective of the Image Sequence Processing Group is to utilize knowledge of the human visual system to develop and explore both models which can serve as the basis for image and image sequence processing and coding techniques, and the techniques themselves.
This work is currently proceeding in two directions:
When applying these approaches to image sequences, it is useful to view the sequence as a spatiotemporal volume of data (as shown above), with motion reflected in the structure of the volume. This treatment of motion is very promising for the efficient representation of motion video, and has many conceptual advantages, as well.
Applications of this work include the development of coding methods which perform well over a broad range of compression ratio and image quality requirements, from the very high compression ratios (and moderate image quality) necessary for videotelephone and multimedia information systems, to the moderate compression ratios required for high quality imaging (HDTV and beyond). Novel image and video processing techniques (including a large class of space- and spatiotemporally-variant filtering methods) are also made possible when using these representations.