A method for processing a multi-view video stream includes receiving the multi-view video stream including a random access picture, the random access picture including a random access slice referring to slices existing at a same time in a different view only, receiving identification information indicating that the received multi-view video stream is a stereo video bitstream, obtaining flag information indicating whether the random access picture is used for inter-view prediction based on the flag information, determining initialization information of a reference picture list of the random access picture based on the flag information, initializing the reference picture list of the random access picture using the initialization information, determining a prediction value of a macroblock in the random access picture based on the initialized reference picture list, and decoding the macroblock using the prediction value of the macroblock.