According to an embodiment, an image decoding method includes: determining at least one prediction unit included in a current coding unit that is one of at least one coding unit splitting the image; generating a merge candidate list related to a current prediction unit included in the at least one prediction unit; determining a prediction mode to be performed in the prediction unit based on the merge candidate list; and performing prediction according to the determined prediction mode, wherein the generating of the merge candidate list further comprises determining whether to add a view synthesis prediction merging candidate to the merge candidate list based on whether a neighboring prediction unit adjacent to the current prediction unit has been encoded using the view synthesis prediction merging candidate.