Intra Prediction Modes for Video Coding Efficiency

The patent segments the 65 intra prediction modes into multiple groups or categories, allowing the decoder to process modes in a structured manner. This segmentation enables efficient mode classification and reduces the computational burden by organizing the large number of modes into manageable sets, thus resolving the contradiction between having many modes for accuracy and maintaining decoder complexity at acceptable levels.

The patent employs preliminary action by pre-defining and signaling a limited set of most probable modes (MPMs) before actual decoding. The decoder uses these pre-signalized MPMs to make initial predictions, and only processes additional modes when necessary. This preliminary setup reduces the immediate processing complexity while still allowing access to the full 65 modes when needed for high accuracy.

The patent applies partial action by using the stronger [1 2 3 4 3 2 1]/16 smoothing filter selectively only for larger block sizes where it provides the most benefit, rather than applying it uniformly to all block sizes. This selective application maintains high prediction quality for large blocks while avoiding unnecessary computational overhead for smaller blocks, thus resolving the contradiction between prediction quality and processing time.

The patent implements local quality by adapting the filter strength and type to the specific block size being processed. Different smoothing filters are applied locally according to the block size characteristics - stronger filters for larger blocks and weaker filters for smaller blocks. This localized approach ensures optimal prediction quality for each block type while minimizing overall processing time.

The patent employs dynamics by making the angle signaling adaptive rather than static. The signaling mechanism dynamically adjusts based on the specific prediction context, block characteristics, and neighboring mode information. This adaptability allows the system to use finer angles and more precise angle representations only when necessary, reducing bitstream overhead while maintaining high intra prediction precision when it provides actual benefit.

The patent uses feedback mechanisms where the encoding process analyzes neighboring block modes and current block characteristics to determine the optimal set of angles to signal. This feedback-driven approach ensures that only the most relevant and useful angle information is transmitted in the bitstream, reducing overhead while maintaining high prediction precision through context-adaptive angle selection.