Assistive visual framing in 3D dense points cloud

Recent progress in digital technology and its popularity in the context of cultural heritage are contributing to the emergence of several applications using a very large and dense digitization of archaeological sites or historical monuments. Navigation in such virtual environment is difficult especially for novice users. In this paper, we propose a new approach in order to make easier the navigation process by performing an assistive visual framing. This approach exploits a new visual feature: the image saliency-based Gaussian mixture. We applied our method on many different environments and we present the example of the Saint Sébastien chapel of the cathedral Notre Dame of Amiens. In order to evaluate our method, we present user evaluation and comparison between manual, assistive and automatic framing.