Error concealment for video transmission using adaptive principal component analysis with missing data

A new adaptive model based algorithm is presented in this paper for real time error concealment application. In proposed algorithm, a modified incremental principal component analysis with missing data is investigated to update the eigenspace during the video process leading to more accurate and efficient eigenspace model. Experimental results demonstrate that proposed algorithm outperforms fixed eigenspace based error concealment and interpolation based error concealment. This improved performance is stable across different quantization levels, loss patterns and loss rates.