Error-Resilient Illumination Compensation in 3D Video Coding

The patent extracts and removes the indirect parsing dependency on reference pictures by introducing a direct ic_flag syntax element in the coding unit header. This flag is independently parsed without relying on reference picture information, thereby eliminating the error propagation path while maintaining illumination compensation functionality.

The ic_flag acts as an intermediary element that mediates between the need for illumination compensation and the requirement for error resilience. By using this explicit flag, the system avoids the harmful indirect dependency on reference pictures while still enabling IC when needed.

The patent applies local quality by making the ic_flag incorporation conditional on the prediction unit type. The flag is only incorporated for 2Nx2N prediction units, which are the most common and benefit most from illumination compensation, rather than forcing it on all coding units. This selective approach maintains error resilience where needed while minimizing bitrate overhead.

The patent uses partial action by applying the error-resilient ic_flag mechanism only to 2Nx2N prediction units rather than all coding units. This partial application achieves the essential error resilience goal while avoiding excessive bitrate increase that would result from universal application.

The patent segments the coding unit based on prediction unit size, applying the ic_flag syntax only to 2Nx2N PUs. This segmentation simplifies the parsing logic by creating a clear distinction between PU types that require the flag and those that don't, thereby reducing overall parsing complexity while maintaining necessary flexibility.

Instead of making the complex case (multiple PU sizes) the default and adding exceptions, the patent inverts the approach by making the simple case (2Nx2N PU with ic_flag) the standard and treating other PU sizes as exceptions that don't require the flag. This inversion simplifies the overall system design.