New sample sets and new down-sampling schemes for linear component sample prediction

摘要

Deriving a linear model for obtaining a sample for a block of a first image component from an associated sample of a second-component block in the same frame comprises determining 1003 two points A, B based on reconstructed samples of both the first-component and the second-component. Each point is defined by two variables, the first variable corresponding to a second-component sample value, the second variable corresponding to a first-component sample value. Parameters of a linear equation representing a straight line passing by the two points are determined 1005 and the linear model defined by the straight-line parameters is derived 1006. Determining the parameters uses integer arithmetic. The method may be used in calculating parameters in cross-component linear mode (CCLM) prediction of luma or chroma values during image compression. The points may be determined based on sample values of pairs of first component and second component blocks neighboring the second component block. A parameter α may correspond to a line’s slope and parameter α may be estimated by calculating a difference between the first components of the points, the size of the difference being restricted by applying a bit shift to it. The bit-shift may depend on a sample bit depth.