Method for generating prediction block in amvp mode

Abstract

A method of decoding an image in a merge mode, the method including deriving, by a motion compensation predictor, motion information using effective spatial and temporal merge candidates of a current prediction unit; generating, by the motion compensation predictor, a prediction block of the current prediction unit using the motion information; entropy decoding, by an entropy decoder, a bitstream to generate a quantized coefficient sequence, inversely scanning, by an inverse quantizer/inverse transformer, the quantized coefficient sequence to generate a quantized block, inversely quantizing, by the inverse quantizer/inverse transformer, the quantized block to generate a transformed block using a quantization parameter, and inversely transforming, by the inverse quantizer/inverse transformer, the transformed block to generate a residual block; and generating, by an adder, a reconstructed block using the prediction block and the residual block, wherein a motion vector of the temporal merge candidate is a motion vector of a temporal merge candidate block within a temporal merge candidate picture, and the quantization parameter is derived using an average of two effective quantization parameters among a left quantization parameter, an upper quantization parameter and a previous quantization parameter of a current coding unit, wherein when the quantized block is larger than a predetermined size, the quantized block is generated with plurality of subblocks, and a scan pattern of inversely scanning for the plurality of subblocks is the same as a scan pattern of inversely scanning for quantized coefficients of each subblock, and wherein the bitstream includes information indicating a position of a last non-zero quantized coefficient in a transform unit.