Video Decoder Reference Picture Buffer Management

The video sequence is segmented into independent picture units (tiles or slices) that can be decoded independently. Each picture is coded with both intra and inter prediction capabilities, allowing the sequence to be divided into self-contained segments that maintain temporal compression while enabling random access to any segment without requiring decoding of previous segments.

Different prediction methods are applied locally to different regions within pictures. Some regions use intra prediction for random access capability, while other regions use inter prediction for better compression efficiency. This local differentiation allows the system to achieve both random access and coding efficiency simultaneously by optimizing each region's prediction method based on its specific requirements.

Reference picture lists are merged across slice boundaries, allowing the same reference picture list to be used for multiple slices. This reduces the complexity of constructing and managing separate reference picture lists for each slice while maintaining decoding accuracy, as the shared reference lists still provide the necessary temporal prediction capability across different slices.