3D Printable Hair Modeling Based On Geometric Texture

The human portrait sculpture is a classical and magnificent artistic medium,it takes real or virtual characters as the basic.An artist usually creates a human figurine using plastic or engraving material,for remembrance or reflection of social life.The large figurines usually become the landmarks of a city or region,while the small figurines are used for decoration and memorial.Due to the high technology difficulty level,the sculpture has always been the artists' patent.While,recent years have shown a rapid development of human character 3D scanning.With the facilities of 3D printing techniques,the 3D printed human figurines become reachable for ordinary users,and the "mass-sculpture" becomes possible.In the literature,more attention has been paid to high-quality facial reconstruction.While,the hair also plays a pivotal role in gaining the realism of the figurines.However,existing hair reconstruction methods either require high-end acquisition equipment or lack geometric details.Moreover,due to the high density,strong randomness,and weak photonasty,hair reconstruction,especially dark hair reconstruction,is particularly challenging.In this paper,we present a novel method for producing high detailed hair using low-cost acquisition equipment for 3D printed human figurines.Our proposed method can reconstruct a closed-manifold surface,make the generated hair appearance similar to the input subject's hairstyle,and reflect personalized details on the hair surface.Our system needs two portrait images,one close-up hair image,and an initial reconstruction model of the subject as the input.The main method is divided into two steps,which concludes coarse template matching and fine details generating.The former guarantees the integrity of the hairstyle,and the latter focuses on the portrayal of the local detail.First,we present a 2D contour matching method,to find the most suitable hairstyle template model from the database to the subject's hair region.Next,we propose a novel fine-level geometric texture generation algorithm,which extracts structure features of the subject's hair from a single image and extends abstract geometric details to the surface of matched template model that is coherent with the personalized hairstyle.Our method fits in traditional pipelines for a figurine reproduction.We prove our method's robustness and effectiveness by capturing several subjects with various hairstyles and 3D printing the figurines.