Motion Vector Predictor Scanning for Video Encoding Efficiency

The patent segments the scanning process by dividing neighboring prediction units into different groups (spatial neighbors and temporal neighbors) and applies different scanning strategies to each group. This segmentation allows the system to reduce the total number of units scanned while maintaining coding efficiency by focusing computational effort on the most promising candidates first.

The patent implements partial action by scanning only a subset of neighboring prediction units rather than all of them. The method scans spatial neighbors first, and only if no suitable predictor is found, then scans temporal neighbors. This partial scanning approach reduces processing load while still achieving adequate coding efficiency by capturing the most significant motion vector predictors.

The patent applies preliminary action by performing a quick scan of spatial neighboring prediction units first, which are typically the most relevant for motion prediction. Only if this preliminary scan fails to find a suitable predictor does the system proceed to scan temporal neighbors. This preliminary scanning approach reduces encoding time by avoiding unnecessary scans while maintaining prediction accuracy through selective scanning.

The patent applies local quality by treating different types of neighboring units (spatial vs. temporal) with different scanning priorities and criteria. Spatial neighbors are scanned first with certain conditions, and temporal neighbors are scanned only under specific circumstances. This localized quality approach ensures that scanning effort is concentrated where it provides the most benefit for prediction accuracy.

The patent segments the motion vector prediction process into distinct stages: first scanning spatial neighboring units, then scanning temporal neighboring units only if needed. This segmentation simplifies the overall algorithm complexity by breaking down the comprehensive scan into manageable, conditional steps rather than requiring a single complex scan of all units.

The patent implements dynamics by making the scanning process adaptive and conditional rather than static and comprehensive. The algorithm dynamically decides whether to scan temporal neighbors based on the results of the spatial neighbor scan. This dynamic approach reduces algorithm complexity by avoiding rigid, comprehensive scanning while maintaining prediction accuracy through flexible, context-dependent scanning.