Hierarchical Epitome Construction for Scalable Video Encoding

The patent segments the video coding process into distinct resolution levels (base layer and enhancement layers), with each layer having its own epitome construction and prediction process. This segmentation allows independent optimization at each level while maintaining overall consistency through the hierarchical structure, resolving the contradiction between serving multiple clients and maintaining information consistency.

The patent performs preliminary action by constructing epitomes at each resolution level before the actual prediction process. These pre-constructed epitomes serve as reliable reference structures that ensure consistency across layers, allowing the system to maintain predictability while adapting to different client requirements through the hierarchical arrangement.

The patent implements a nested structure where lower-resolution epitomes are embedded within higher-resolution epitomes in a hierarchical arrangement. This nesting allows prediction quality to improve at higher resolutions while maintaining data efficiency, as the lower-resolution information is reused and refined rather than completely replaced, thus resolving the contradiction between prediction quality and data volume.

The patent applies local quality by allowing different resolution levels to have different levels of detail and complexity appropriate to their specific requirements. Base layers use coarser epitomes for efficient transmission, while enhancement layers use finer epitomes for improved prediction quality, optimizing the balance between data volume and prediction accuracy at each local level.