Accurate silhouette extraction of multiple moving objects for free viewpoint sports video synthesis

In this paper, we propose a new method of automatic silhouette extraction of multiple moving objects with high accuracy for free viewpoint stadium sports video synthesis. The proposed method is basically composed of three parts, including a global extraction based on temporal background subtraction, a classification step based on the constraints of extracted candidates of objects, and local refinement based on the statistical information of the chrominance component of each extracted object. Experimental results show that the proposal outperforms the temporal background subtraction model and Gaussian Mixture Model (GMM) in terms of both objective and subjective evaluations. In addition, the quality of the synthesized free viewpoint sports video is also enhanced by adopting more accurate silhouettes of objects that are extracted by our proposed method. Furthermore, as there is no manual operation in the proposed method, the automatic multiple silhouettes extraction has also been fully realized.