Face recognition based on depth maps and surface curvature

摘要: This paper explores the representation of human face by features based on curvature surface. Curature captures many necessary to accurately describe face, such as shape forehead, jawline, and cheeks, which are not easily detected from standard intensity images. Moreover, value at a point surface is also viewpoint invariant. Until recently range data high enough resolution accuracy perform useful calculations scale had been unavailable. Although several researchers have worked problem interpreting curved (although usually highly geometrically structured) surfaces, main approaches centered segmentation signs mean Gaussian proved sufficient in themselves for case face. details calculation principal particular set, general descriptors curvature, specific both priori knowledge about structure These can be incorporated into different recognition strategies. A system that implements one strategy, depth template comparison, giving rates between 80% 90% described.