Mutual Information in 3D Video

3D video, which is captured by multiple synchronized cameras and consists of a sequence of mesh models, is becoming a hot topic recently. Similarity measure between frames is a powerful tool in video analysis. In this paper, an effective similarity measure based on mutual information is proposed for 3D video. Only vertex positions in mesh models are utilized in our mutual information. The mutual information is calculated based on the distance field in volumetric models, which are converted from mesh models. Our experiments demonstrate the mutual information can reflect the motion of the object in 3D video very well as a similarity measure.