Viewpoint Scoring Approach and Its Application to Locating Canonical Viewpoint for 3D Visualization

In this paper, a novel viewpoint scoring method based on multi-attribute fusion is proposed. The perceptual model of viewpoint preference is explored from geometry and visual perception aspects. Modified mesh saliency entropy is presented as the crucial intrinsic geometric attribute. Several digital image factors which have influence on human visual perception form the viewpoint perception attributes. Evolution algorithm is utilized to select the canonical viewpoint automatically and intelligently. Experimental results demonstrate that the canonical viewpoint obtained by the proposed method contains more visible salient features and better conforms to human visual perception characteristic. Moreover, the method has high efficiency and requires no user interaction.