Delay-constrained three-dimensional graphics streaming over lossy networks

The objective of this thesis is to deliver high-quality 3-D graphics to the clients with minimum end-to-end delay. This research proposes an end-to-end 3-D streaming system that is scalable with respect to network behavior. Current multimedia streaming techniques are limited to voice, image or video sequences. 3-D models have different content properties and therefore new streaming techniques need to be developed that are specially tailored for 3-D models. Similarly, current 3-D streaming systems are limited to progressive compression and data partitioning and these techniques are not scalable with respect to channel behavior. The proposed end-to-end transmission system combines (i) 3-D model properties, (ii) application-layer functionalities such as forward error correction (FEC) codes and multiple description codes (MDC), and (iii) network-layer functionalities such as re-transmission in order to minimize the expected the distortion in the decoded model or to minimize the end-to-end delay.; The proposed research achieves error-resiliency by using unequal error protection. Some parts of the transmitted bitstream are more important than other parts and hence more error-protection bits are assigned to these important segments of the bitstream. Preliminary results show a graceful degradation in the decoded 3-D model quality, as the packet loss rate increases, compared to treating all parts of the bitstream equally. We further propose a bit-allocation algorithm that distributes the bit budget between geometry precision, tessellation density, and error protection bits. Finally, we propose a novel transport protocol especially designed for 3-D models. We call this protocol the 3-D models transport protocol (3TP). In 3TP, we intelligently utilize the transport control protocol (TCP) and the user datagram protocol (UDP) to transmit different parts of the bitstream in order to meet certain distortion constraint while minimizing the end-to-end delay.