High-definition texture reconstruction for 3D image-based modeling

Abstract

Image-based modeling is becoming increasingly popular as a means to create realistic 3D digital models of real-world objects. Applications range from games and e-commerce to virtual worlds and 3D printing. Most research in computer vision has concentrated on the precise reconstruction of geometry. However, in order to improve realism and enable use in professional production pipelines digital models need a high-resolution texture map. In this paper we present a novel system for creating detailed texture maps from a set of input images and estimated 3D geometry. The solution uses a mesh segmentation and charting approach in order to create a low-distortion mesh parameterization suitable for objects of arbitrary genus. Texture maps for each mesh segment are created by back-projecting the best-fitting input images onto each surface segment, and smoothly fusing them together using graph-cut techniques. We investigate the effect of different input parameters, and present results obtained for reconstructing a variety of different 3D objects from input images acquired using an unconstrained and uncalibrated camera.