We have developed a Gaussian mixture model-based technique for the compensation of lost pixel blocks during real time transmission of video. In this paper, we pursue an argument in order to better understand how the Gaussian mixture model-based estimator works. The discussion is supported with subjective evaluations and examples. Naive viewers preferred the result of our proposed technique 72 percent of the time in comparison with the linear minimum mean square error estimator. Our scheme increases performance measured in peak signal-to-noise ratio for all the 11 standard evaluation movie clips that were used.
展开▼
