Fast Evaluation of Packet Loss Impairment on MPEG-2 Streaming Video Using Fisher Classifier

Compressed videos are extremely sensitive to packet loss in today´s best effort IP network. Accurate detection and evaluation of packet loss impairment on streaming video is thus very important. In this paper, we propose two algorithms combining together to effectively detect and evaluate these impairments. The basic idea of our algorithm is to discriminate impaired frames and unimpaired frames in pixel level based on simple features of hue and intensity of impairments. We have designed a Fisher classifier using data collected in our lab. The algorithm performs well in our experiments and has low computational complexity. Our algorithms are currently being implemented in the quality monitoring component of our HDV over IP system. It could also be used in adaptive system or be used to compare different error concealment strategies