A comparison between two 3D free-viewpoint generation methods: Player-billboard and 3D reconstruction
Tetsuya Shin, Nozomu Kasuya, Itaru Kitahara, Yoshinari Kameda, Yuichi Ohta · 2010
This paper investigates the optimal 3D modeling solution for making free-viewpoint video in a soccer stadium. We compare a player-billboard method and a 3D reconstructing method that exploits a shape-from-silhouette approach. To examine the influence of noise and the number of cameras used to make the free-viewpoint video, we produce a CG simulation of a soccer player in action and conduct subject-based evaluation in a pair-wise comparison of the videos made by these two methods. The evaluation is made under the method of limits. From the results of the evaluation, the player-billboard method is more robust to noise than the shape-from-silhouette method, and it also better represents smooth motion parallax as the number of capturing cameras is increased, although representing parallax has traditionally been a weak point of the player-billboard method. Our results show that the player-billboard method is superior to the shape-from-silhouette method for generating 3D free-viewpoint video.