Programmable Spatial Prediction for Video Encoding

The patent implements dynamic spatial prediction by allowing the encoder to select from multiple prediction modes (planar, differential, gradient) and adjust prediction parameters based on the actual video content characteristics. The encoder dynamically determines the optimal prediction mode for each block by evaluating gradient calculations and comparing prediction errors, rather than using a fixed prediction approach. This dynamic adaptation enables better bit rate reduction efficiency while managing complexity through standardized algorithms.

The patent changes prediction parameters dynamically by calculating gradients in multiple directions (horizontal, vertical, diagonal) and selecting the prediction mode that provides the best match for each video block. The encoder computes gradient values, determines optimal prediction directions, and adjusts prediction parameters based on content characteristics such as edge orientations and texture patterns. This parameter adaptation significantly improves prediction accuracy and perceptual quality while maintaining encoding simplicity through standardized decision algorithms.

The patent segments the video content into discrete blocks and applies appropriate prediction modes to each block independently. By dividing the video stream into manageable units (macroblocks and sub-blocks), the encoder can efficiently process each segment using optimized algorithms for that specific region. This segmentation enables parallel processing capabilities and allows the use of computationally intensive gradient calculations only where necessary, rather than uniformly across the entire video stream, thus maintaining high encoding speed while achieving effective bit rate reduction.