Tampere Finland

Video communication systems can often make use of unequal error protection (UEP). In order to apply UEP, video bit-streams have to be organized in portions of different importance in terms of visual quality. Techniques achieving this goal include data partitioning and scalable video coding. These techniques generally treat an entire image equally in spatial domain, while many images have distinct spatial regions of interest. It is intuitively clear that these regions of interest should have better error protection than other areas. No simple method to enable separation of interest regions has been adopted to the ITU-T video coding standards so far. This contribution proposes image segmentation to one or more rectangular foreground sub-pictures and to a hollow background sub-picture. Sub-picture boundaries follow macroblock boundaries. The boundaries are treated as slice boundaries or as picture boundaries depending on the signaled coding mode. Due to the rectangular shape of the foreground sub-pictures, implementation of the proposed scheme is straightforward and practically no additional computations are required in a decoder implementation compared to current TML.