The method includes the steps of receiving the multiview video data stream including a random access picture including a random access slice, the random access slice referencing only slice corresponding to a same time and a different view of the random access picture, obtaining initialization information of a reference picture list for the random access slice, the initialization information representing a reference relation between a plurality of views with view number information and view identification information, initializing the reference picture list for inter-view prediction by appending an inter-view reference index to an initialized reference picture list for temporal prediction, the inter-view reference index being appended based on the view number information and the view identification information, obtaining a difference value from the a multi-view video data stream according to the type information, the difference value representing a residual of inter-view reference index in the initialized reference picture list; determining an assignment modification value for modifying the inter-view reference index in the initialized reference picture list according to the difference value, modifying the initialized reference picture list for inter-view prediction using the determined assignment modification value, determining a prediction value of a macroblock in the random access picture based on the modified reference picture list, and decoding the macroblock using the prediction value, wherein the initialization information is obtained from an extension area of a sequence header.