Internetworking and Media Communications Research Laboratories
Department of Computer Science,
DYNAMIC GAZE SPAN WINDOW BASED FOVEATION FOR PERCEPTUAL
Prof. Javed I. Khan
Date Submitted: May 2003
Human vision provides numerous opportunities for video data-compression. Human vision extends about 140 degree, but only about 2 degrees have sharp vision. A fascinating body of research exists in vision and psychology geared towards the understanding of the human visual perception system. This thesis presents a novel eye-gaze enhanced media transcoding system for streaming video. This scheme includes a video server, a real-time performance capable media transcoder, a video player and an eye tracker. The system intakes live perceptual information related to a subject’s eye position. Eye and head movements are detected via an eye-tracker, and a magnetic head tracker. A unique challenge of this real time perceptual adaptation scheme is the incorporation of fast eye movement mechanisms into a complex MPEG-2 transcoding scheme. An important factor in this perceptually adaptive encoding method is the delay between the time an eye-gaze sample is taken and the time the coding response arrives on the screen. This delay is particularly critical if the video involves network transmission. The delay also usually increases when large format media is to be perceptually transformed due to the coding complexity. This thesis investigates this feedback delay compensation problem and proposes a novel gaze interaction based foveation windowing scheme to solve it. The proposed technique is able to contain 90% of the gazes within 20-25% window coverage area. The media transcoder developed on this scheme is one of the first eye-gaze based perceptual transcoders. It can be used between the server and the video player in a networked environment. The architecture of the transcoder is designed to allow transmission of both stored and live media. Though the architecture is independent of any media type, this system currently handles ISO/IEC 13818-2 MPEG-2 standard.
· “Dynamic Reflex Window Prediction for Perceptual
Coding of Wide Format Video Stream With Eye Gaze Tracking,” Javed Khan,
Oleg Komogortsev. Proceedings of Circuits, Signal and Systems conference,
· “Contour Approximation can Lead to Faster Object Based Transcoding with Higher Perceptual Quality,” [technical report: January 2003].
· “Dynamic Gaze Span Window based Foveation for Perceptual Media Streaming,” [technical report: November 2002].
· “Perceptually Encoded Video Set from Dynamic Reflex Windowing,” [technical report: July 2002].
Adaptive Netcentric Systems on Active Network: A Self-Organizing Video Stream
that Auto Morphs Itself while in Transit via a Quasi-Active Network,”
“Resource Adaptive Netcentric
Systems: A Case Study with SONET – a Self – Organizing Network
Embedded Transcoder.” Javed I. Khan, Seung S. Yang, Qiong Gu, Darshan
Patel, Patrik Mail, Oleg Komogortsev, Wansik Oh, & Zhong Guo. Proceedings
of the ACM International Conference on Multimedia, ACM MM’2001,
· ‘Impact on Stream Compression and Video Quality Motion Vector Reuse Transcoding” [technical report: February 2002]
Last Modified: May 2003.