Fast Texture Mapping Adjustment via Local/Global Optimization

This paper deals with the texture mapping of a triangular mesh model given a set of calibrated images. Different from the traditional approach of applying projective texture mapping with model parameterizations, we develop an image-space texture optimization scheme that aims to reduce visible seams or misalignment at texture or depth boundaries. Our novel scheme starts with an efficient local (and parallel) texture adjustment scheme at these boundaries, followed by a global correction step to rectify potential texture distortions caused by the local movement. Our phased optimization scheme achieves 50~100 times speed up on GPU (or 6× on CPU) compared to previous state-of-the-art methods. Experiments on a variety of models showed that we achieve this significant speedup without sacrificing texture quality. Our approach significantly improves resilience to modeling and calibration errors, thereby allowing fast and fully automatic creation of textured models using commodity depth sensors by untrained users.

1 Introduction

With the wide availability of low cost commodity 3D full frame sensors, such as the Microsoft Kinect and Intel Real-Sense, it is becoming easier and easier to capture 3D data. Starting from the robust system of KinectFusion [1], [2], large scale scanning [3], [4], to the more recent work of fusion of dynamic objects over space and time [5], [6], good quality 3D models can be obtained even with a single hand-held depth camera. These approaches typically use a volumetric representation of 3D geometry, the final model is mostly presented as a white (texture-less) model or colored with simple per-vertex color. Given the high cost associated with a dense voxel grid, per-vertex colored model appears to be blurry and low resolution, the final model's appearance captured by these methods left something to be desired.