This paper proposes a new hole filling method using 3D geometric transformation to make a synthesized image at virtual viewpoint. Disoccluded regions in virtual viewpoint could appear as hole regions in the synthesized image. The hole regions are supposed to be filled in a visually plausible manner. In most of previous works, hole regions were filled with translated patches (labels) extracted from non-hole region (source region). In many cases, however, it is difficult to find the most similar labels due to texture irregularity and perspective distortion in a wide view angle. In this paper, we propose a new hole filling method in which hole regions are filled with the best labels using 3D geometric transformation and associated nearest neighbor search. Experimental results show that the proposed method provides structurally consistent results with a high computational efficiency, which outperforms existing hole filling methods.1