A statistical error detection technique for low bit-rate video

Multimedia services requiring image and video compression are now widely available over fixed telecommunications networks such as cable television networks and the Internet. In the future, these services will also be provided over wireless networks similar to those currently used for digital mobile telephony. As a consequence of the high bit error rates associated with these channels, special error-resilient coding techniques are required to achieve a satisfactory picture quality. An error detection scheme based on the detection of statistically incorrect information, is introduced. The new technique calculates statistical information about the video sequence currently being decoded and uses this information to detect and conceal macroblocks which have been incorrectly decoded. Simulation results show this technique provides a significant improvement in the decoded image quality while requiring no additional information to be transmitted to the decoder.