The　goal　of　casual　stereoscopic　photography　is　to　allow　ordinary　users　to　create　a　stereoscopic　photo　using　two　photos　taken　casually　by　a　monocular　camera.　In　this　poster,　we　propose　a　multiagent　reinforcement　learning　framework　for　visual　comfort　enhancement　of　casual　stereoscopic　photography.　Each　agent　calculates　a　homography　matrix　based　on　the　positions　of　four　comers　before　and　after　the　offsets.　Furthermore,　we　propose　a　hierarchical　stereo　transformer　based　on　window　attention,　which　enhances　and　fuses　multiscale　correlations　between　left　and　right　views,　Experimental　results　show　that　our　proposed　method　achieves　superior　performance　to　the　state-of-the-art　methods.