METHOD FOR ENCODING AND/OR DECODING IMAGES ON MACROBLOCK LEVEL USING INTRA-PREDICTION

摘要

Conventional intra-prediction uses pixels from left and upper neighbour blocks to predict a macroblock (MB). Thus, the MBs must be sequentially processed, since reconstructed left and upper MBs must be available for prediction. In an improved method for encoding Intra predicted MBs, a MB is encoded in two steps: first, a first portion of the MB is encoded independently, without references outside the MB. Pixels of the first portion can be Intra predicted using DC mode. Then, the first portion is reconstructed. The remaining pixels of the MB, being a second portion, are intra predicted from the reconstructed pixels of the first portion and then reconstructed. The first portion comprises at least one column or one row of pixels of the MB. The encoding is applied to at least two Intra predicted MBs per slice, or per picture if no slices are used.

主权项

1. A method for encoding macroblocks of an image using intra prediction, comprising steps of
dividing a macroblock into two portions, wherein a first portion comprises at least one column or one row, and a second portion comprises remaining pixels of the macroblock; encoding pixels of the first portion of the macroblock without referencing pixels outside the first portion of the macroblock; and encoding pixels of the second portion of the macroblock using intra prediction, wherein the intra prediction references said pixels of the first portion of the macroblock or previously encoded pixels of the second portion of the macroblock.