DCT coefficient-based error detection technique for compressed video stream

In the process of transmission of digital encoded video bit streams over a physical network, the data is corrupted by random bit errors, which has a severe effect on the decoded video quality. The success of an error concealment algorithm depends heavily on the detection of the corrupted regions. We propose a new error detection algorithm for discrete cosine transform based video coding. Our algorithm detects thus far undetectable erroneous regions with a very little increase in computational complexity. We show that erroneous blocks in the decoded frames can be detected by monitoring the variation of average energy from block to block. To avoid false detection due to the presence of strong edges we employ a very fast DCT coefficient based edge detection algorithm. Computer simulation shows that the quality of the recovered image is significantly improved in comparison to well-known methods.