Disclosed herein is an image encoding/decoding method. The method of decoding the image includes structuring a motion information candidate list of a current block, selecting a first motion information candidate used for prediction of a first subblock in the current block from the motion information candidate list, selecting a second motion information candidate used for prediction of a second subblock in the current block from the motion information candidate list, generating a prediction sample of the first subblock by performing inter prediction with respect to the first subblock based on the first motion information candidate, and generating a prediction sample of the second subblock by performing inter prediction with the second subblock based on the second motion information candidate. The first motion information candidate is any one of candidates in a first prediction direction in the motion information candidate list, and the second motion information candidate is any one of candidates in a second prediction direction in the motion information candidate list.