Best view methods suitability for different types of objects

Abstract

In this paper, we present three approaches to the best view selection for 3D objects, in order to find the proper combination. The first method is based on geometry and uses the maximum amount of visible vertices. The second method uses the visual attention model and the third one calculates the amount of information in a scene after projecting the 3D scene to the 2D image using entropy. Several 3D models from Blender are used in the experiment. We compare these different approaches and explore their suitability on various objects. An online questionnaire is made to find out which of the tested images fits better to the human tastes.