Cross-Component Prediction Candidate Selection for Video Blocks

The patent performs preliminary classification of neighboring blocks to identify suitable candidates for cross-component prediction before actual encoding. By pre-processing and organizing candidate blocks in advance, the method improves both coding efficiency and effectiveness without requiring complex real-time computations during the main encoding process.

The patent introduces new parameters and criteria for evaluating neighboring blocks, such as block type classifications, gradient calculations, and similarity metrics. These parameter changes enable more sophisticated selection of prediction candidates, directly addressing the insufficiency of existing standards while maintaining computational feasibility.

The patent segments the candidate selection process into distinct stages: identifying neighboring blocks, classifying block types, calculating evaluation metrics, and selecting final candidates. This segmentation reduces processing complexity by breaking down the complex task into manageable steps, while systematically improving coding effectiveness through thorough evaluation.

The patent applies different evaluation criteria and selection strategies for different types of neighboring blocks (e.g., intra-predicted blocks, inter-predicted blocks, blocks with different gradients). This local quality approach ensures that each block type is evaluated using the most appropriate metrics, improving overall coding effectiveness without uniformly increasing processing complexity across all blocks.