Spatial temporal concealment of lost blocks in coded video

When errors are introduced into a compressed video bit-stream, blocks of pixels are lost. The decoder attempts to conceal this loss to maximize the displayed picture quality. Previous approaches use either surrounding pixels in the current picture to conceal lost blocks (spatial concealment) or pixels from the previous picture (temporal concealment). We show how joint spatial temporal concealment can be implemented to provide higher quality concealment. To do that we model video sequences as mixtures of simple Markov processes and utilise the model for concealment of portions of video which are missing. This model allows us to combine spatial and temporal information in a seamless manner when performing concealment on the missing blocks.