Primary transforms for cross-component level reconstruction
摘要
說明書
In some example implementations, when an EOB for signaling a relative end-of-block position associated with a transform block of a second color component is zero, which indicates that all transform coefficients in this transform block are zero and if CCLR is applied on this transform block, then a CCLR refinement process may be applied on each transform coefficient in this transform block, for example, by adding an offset value corresponding to the each transform coefficient. The offset value may be derived base on a co-located transform coefficient in a transform block of a first color component. As a result of the refinement process, the refined transform coefficients in the transform block are no longer all zero. Following the refinement process, an inverse transform may be performed on the CCLR refined transform block.
In some example implementations, rather than targeting the whole transform block, the CCLR refinement process may only target a portion of the transform coefficients in the transform block.
In some example implementations, when applying the inverse transform on the CCLR refined transform block of the second color component, the same transform kernel as used for the inverse transform on the co-located transform block of the first color component may be selected. Note that the co-located transform block of the first color component is used as a basis (or reference) for deriving the offset values used in the CCLR refinement process which is applied on the transform block of the second color component.
