A method for decoding a multilayer video signal, according to the present invention, is characterized by: selecting, from a corresponding picture of at least one reference layer, a candidate reference picture of a current picture by using sublayer number information relating to the reference layer and a temporal ID of the current picture belonging to a current layer, determining the number of active references for a current picture based on the number of the candidate reference picture, acquiring a reference layer Identifier (ID) based on the determined number of active references, determining an active reference picture for the current picture using the reference layer ID, generating a reference picture list for the current picture, the reference picture list including a temporal reference picture and the active reference picture, and performing inter-layer prediction for the current picture based on the reference picture list.