Multimedia Frame Prediction Using Global and Local Reference Blocks

The patent segments the reference data blocks into two distinct sets: a global reference data block set containing non-adjacent coded data blocks, and a local reference data block set containing adjacent coded data blocks. This segmentation allows the system to manage reference data in an organized manner, improving coding performance by providing diverse reference options while maintaining manageable complexity through structured classification.

The patent introduces adaptive selection mechanisms that dynamically choose which reference data block sets to use based on the specific coding requirements and characteristics of each data block. This dynamic approach enables the system to optimize coding performance for different scenarios while avoiding the need to always maintain and process the full complex reference structure, thus resolving the contradiction between performance and complexity.

By dividing reference data blocks into global and local sets with clear definitions (non-adjacent vs. adjacent), the patent enables faster selection processes. The segmented structure allows the system to quickly determine which set or combination of sets is appropriate for each coding task, reducing search time while maintaining access to a comprehensive reference library for optimal performance.

The patent performs preliminary organization and classification of reference data blocks into structured sets during the preparation phase. This preliminary action of categorizing and pre-organizing reference data enables rapid retrieval and selection during the actual coding process, reducing the time required to process multiple reference options while preserving the benefits of having an extensive reference library.

The patent segments the adaptive selection process into clear, manageable categories by dividing reference data blocks into global and local sets. This segmentation provides a structured framework for adaptation that is easier to implement and manage, allowing the system to handle different application scenes through a organized classification system rather than a complex monolithic algorithm.

The patent applies local quality by creating locally relevant reference data block sets (local reference data block set with adjacent blocks) alongside globally relevant sets. This allows the system to adapt to specific local requirements of different data blocks while maintaining access to global patterns, achieving versatility through a differentiated reference structure that balances local and global considerations.