Predictive decoding for delay reduction in video communications

Low delay is critically important for interactive video communication. Unpredictable delays and bursty traffic in today’s networks may significantly degrade the performance of interactive video services. This thesis presents several predictive decoding techniques for delay reduction. The basic idea is to predict future video frames from past video data, and display them before they arrive at the decoder. Inevitably, this will reduce the quality of the displayed frames somewhat, but it will also enable the user to choose the proper trade-off between quality and delay. The frame prediction module was implemented as an add-on to MPEG-4 video decoder, and tested on a variety of standard sequences. The performance highly depends on the characteristics of the sequence, such as motion intensity, frame rate, etc. Our results indicate that in most cases, it is possible to reduce the perceived end-to-end communication delay by about 100 ms while maintaining reasonable video quality.