Video Prediction Filtering for Higher Coding Efficiency

The patent applies filter-based prediction and reconstruction refinement methods that modify the parameters of video processing operations. By applying filters to prediction samples and reconstructed video units, the patent changes the quality parameters of the coding process, achieving improved prediction accuracy and reconstruction quality without adopting entirely new coding frameworks, thus resolving the contradiction between maintaining compatibility and improving coding efficiency.

The patent performs filtering operations on prediction samples and reconstructed video units during the encoding and decoding processes. By applying filters in advance to prediction samples before generating the final prediction, and to reconstructed video units before output, the patent improves coding efficiency through preliminary processing steps that enhance quality without requiring fundamentally more complex coding architectures.