FIR-based interpolation in advanced video codecs on VLIW processor

摘要

A half sample and quarter sample pixel value interpolation calculation classifies a set of all possible half sample and quarter sample pixel value interpolation calculations into a plurality of interpolation types. A function kernel for each interpolation type includes a half sample pixel and all quarter sample pixels whose value calculation depend on a value of said half sample pixel value. The correct function kernel is called invoking the corresponding value interpolation calculation. The function kernels operate iteratively over a plurality of macroblocks of at least two macroblock sub-partition sizes. The calculation of dependent quarter sample pixel values preferably employs a packed average instruction.