Asymmetrical Deblocking Filter for Video Coding Error Reduction

The patent applies asymmetry by determining different filter characteristics for pixels on opposite sides of block boundaries. Specifically, when pixel error distribution is discontinuous, the encoder applies stronger filtering to pixels with larger errors and weaker filtering to pixels with smaller errors, rather than using uniform symmetrical filters. This resolves the contradiction by sacrificing filtering uniformity to achieve superior error reduction efficiency.

The patent implements local quality by adapting filter characteristics to local error conditions. The encoder evaluates pixel errors individually or in groups and applies customized filter parameters (such as different beta and tau values) to different pixel locations based on their specific error magnitudes. This allows the filtering process to optimize error reduction locally while maintaining overall boundary continuity.

The patent utilizes parameter changes by dynamically adjusting filter parameters (beta and tau values) based on pixel error magnitudes. The encoder determines asymmetrical filter characteristics by modifying these parameters according to local error conditions, allowing efficient adaptation without requiring completely different filter designs. This approach balances improved error reduction with manageable computational complexity through parameter optimization.

The patent applies partial action by selectively applying asymmetrical filtering only where needed. The encoder evaluates error distribution and applies full asymmetrical processing only to pixels with discontinuous error patterns, while using simpler symmetrical filtering for uniform error regions. This selective approach achieves significant error reduction improvement without incurring the full computational cost of asymmetrical filtering everywhere.

The patent implements dynamics by making filter characteristics adaptive rather than static. The encoder dynamically adjusts filter strength based on local error conditions, applying stronger filtering where errors are large and weaker filtering where errors are small. This dynamic adaptation ensures that filtering always operates at the optimal strength for the local conditions, preventing both insufficient error reduction and excessive smoothing that would degrade image quality.

The patent employs feedback mechanisms by using pixel error information to control filter application. The encoder evaluates pixel errors and uses this feedback to determine appropriate filter characteristics for each pixel or pixel group. This closed-loop approach ensures that filtering intensity is continuously adjusted based on actual error conditions, optimizing the balance between error reduction and image quality preservation.