An intra prediction method performed by a decoding apparatus is disclosed, the method comprising the steps of:
deriving a first candidate intra prediction mode based on a left neighboring block of a current block;
deriving a second candidate intra prediction mode based on an upper neighboring block of the current block;
determining an intra prediction mode for the current block as one of candidate modes including the first candidate intra prediction mode and the second candidate intra prediction mode; and
generating prediction samples on the current block based on the determined intra prediction mode,
wherein when the upper neighboring block is not located in a current largest coding unit, LCU, in which the current block is located, the second candidate intra prediction mode is set equal to non-angular mode.