Region-based coding schemes for video sequences have recently received
much attention owing to their potential to enable several interesting
multimedia applications. The authors focus on the optimisation of the
coding of uncovered regions. It is found that the widely-used tools f
or coding the texture within arbitrarily shaped regions of the intra i
mage are not appropriate for failure regions. A new method for doing s
o is proposed which consists of applying a predictive coding step to t
he newly visible portions based on already transmitted spatial data.