Video Decoder Deformation Compensation for Multiview Accuracy

The patent applies deformation compensation in advance during the decoding process by warping reference pictures from different views to match the current picture geometry. This preliminary geometric transformation ensures that corresponding blocks align properly before prediction operations, thereby improving decoding accuracy without requiring complex real-time adjustments during prediction.

The patent introduces deformation compensation parameters as an intermediary element that mediates between reference pictures from different views and the current picture. These parameters (including warp coefficients and block offsets) act as a bridge to transform reference blocks into the correct geometric configuration, enabling accurate prediction while maintaining systematic processing.

The patent merges the deformation compensation parameter calculation process with neighboring block information. By deriving parameters from adjacent blocks rather than calculating them independently for each block, the method reduces redundant computations while maintaining prediction accuracy. This combining approach accelerates the decoding process by eliminating repeated calculations of similar geometric transformations.

The patent applies partial deformation compensation by using parameters from neighboring blocks for the current block. Instead of performing full independent calculations for every block, the method selectively reuses parameters from adjacent blocks where geometric characteristics are likely similar, thereby reducing overall computational complexity while maintaining sufficient prediction accuracy for most blocks.