Disclosed are a candidate mode queue processing method and device thereof. The processing method comprises: selecting an adjacent image block from adjacent image blocks of a current image block and only conducting selection once, and inserting an encoding mode, i.e. a viewpoint synthesis prediction mode, of the selected adjacent image block into a candidate mode queue. The present invention solves the problem in the prior art that the encoding complexity is relatively high when a plurality of redundant viewpoint synthesis prediction candidate modes appear in the candidate mode queue at the same time.