Contrast-based surface saliency

Abstract

The detection of salient regions is an important preprocessing step for the analysis of mesh surfaces. The detected salient region is a reflection of perception-based regional importance for surfaces. It finds many 3D applications, such as mesh simplification, registration, segmentation and compression. In this paper we propose a novel method for the detection of saliency in a 3D surface. Our method incorporates the bilateral normal filtering, shape index and Retinex to generate the surface contrast, then produces vertex-based and region-based contrast saliencies. The effectiveness of this method is demonstrated by visual observation of detected salient regions. Also, the applications of saliency-guided interest points detection and saliency-guided simplification are demonstrated to validate the proposed method. In addition, the criterion to validate the proposed method is the repeatability of the salient points and surface simplificaition errors. A large number of the experiments have been performed over real data and the results demonstrate that the proposed approach has achieved better results than competitors.