Affine Motion Vector Control in Video Encoders for Lower Memory Access

The patent applies parameter changes by modifying motion vectors based on gradient information. Specifically, motion vectors are adjusted according to the gradient magnitude and direction calculated from pixel differences, which changes the motion vector parameters to better match actual motion patterns. This reduces the variation range of motion vectors within blocks, thereby reducing memory access size while maintaining coding efficiency.

The patent implements dynamics by adaptively adjusting motion vectors based on local gradient characteristics. The motion compensation process dynamically modifies motion vectors according to the calculated gradients, allowing the system to adapt to varying motion patterns within different regions of the video block. This dynamic adjustment optimizes both memory access requirements and coding efficiency.

The patent applies segmentation by dividing the video block into smaller regions and calculating gradients separately for each region. This allows the complex gradient-based motion vector modification to be applied locally rather than globally, reducing overall computational complexity while maintaining prediction accuracy in each segmented region.

The patent implements partial action by applying gradient-based motion vector modification selectively rather than to all motion vectors uniformly. The modification is applied based on gradient thresholds and local characteristics, performing the computationally intensive operation only where necessary to improve prediction accuracy, thus balancing accuracy gains with computational cost.