Chroma Prediction Linear Model Using Luma Gradients for Video Coding

The patent changes the parameters used in chroma prediction by introducing gradient values of luma components and neighboring sample values into the linear model. This transforms the prediction from a simple linear relationship to a more complex model that incorporates spatial frequency information and contextual data, thereby improving both prediction accuracy and coding efficiency

The patent creates a composite prediction model that combines multiple factors: the original luma component values, gradient values derived from luma gradients, and neighboring chroma sample values. This composite approach integrates multiple data sources to achieve more accurate chroma prediction while maintaining computational efficiency

The patent segments the prediction process into distinct components: gradient calculation from luma samples, linear combination with chroma samples, and residual handling. This segmentation allows each component to be optimized independently while maintaining overall model manageability and computational efficiency

The patent introduces gradient values as an intermediary element that bridges the luma component and chroma component prediction. The gradient acts as a mediator that captures spatial frequency information and transfers it to the chroma prediction, improving accuracy without directly increasing model complexity