A robust iterative algorithm for reconstruction from redundant filter banks [image coding example]

Because of their interactive nature, multimedia streams must be sent over UDP and suitable countermeasures for minimizing the effect of data loss have to be taken. Among the various proposed techniques, coding with redundant filter banks has been proposed as a mean to add robustness to the data stream. In order to avoid the computationally expensive matrix inversion necessary for signal reconstruction in the presence of packet loss, an iterative reconstruction algorithm can be used. Unfortunately, the classical algorithm for iterative reconstruction does not necessarily converge if too many coefficients are lost. In this paper we propose a modified version which converges even in the case of excessive losses.