US20260222564
2026-07-30
Electricity
H04N19/117
The cross-component sample offset (CCSO) filter is a crucial in-loop filtering tool in the AV2 video coding standard, designed to enhance video quality by correcting reconstruction errors. However, deriving CCSO filtering parameters at the frame level involves a complex, pseudo-brute-force search through numerous parameter combinations, which is impractical for real-world encoder and hardware applications. A hierarchical decision framework has been developed to address this issue by selectively skipping or narrowing the parameter search based on lightweight, content-adaptive analysis conducted beforehand.
This framework uses a decision tree to determine whether to perform a full, truncated, or no search for CCSO parameters, based on frame type and encoded size. By analyzing the frame type and size, the decision tree can either initiate a comprehensive CCSO parameter search, conduct a limited search using pre-determined reference parameters, or skip the CCSO filter entirely. This approach significantly reduces the complexity of the search process while maintaining video quality.
Intra-frames and anchor frames, which are critical reference points for subsequent frames, undergo an adaptive CCSO parameter search due to their importance. This search is more complex but necessary for maintaining quality across frames. For other frames, such as inter-frames, the decision tree uses the encoded bitstream size of the nearest lower-layer reference frame as a proxy for the frame's complexity and information content. This proxy helps decide the appropriate level of CCSO parameter search or whether to skip it.
The decision tree compares the encoded frame size against predefined thresholds to determine the search method. If the frame size exceeds a higher threshold, it triggers an adaptive CCSO parameter search. If the size is between two thresholds, a truncated search is performed using existing reference parameters. If the frame size is below a lower threshold, CCSO filtering is disabled for that frame, maximizing complexity reduction with minimal quality loss.
By implementing this hierarchical decision framework, the complexity of deriving CCSO parameters is significantly reduced, making it feasible for practical encoder deployments and hardware implementations. This method ensures that the CCSO filter is applied efficiently, maintaining video quality while optimizing computational resources. The framework adapts to varying scene complexities and frame types, ensuring that video encoding remains both effective and efficient.