Simultaneous Incremental Reconstruction of Object Geometry and Appearance for Interactive 3-D Model Acquisition

Abstract

While the creation of three-dimensional models from real-life objects is a commonly applied process, the reconstruction of complete datasets from such objects is still a delicate task. In this paper, a simple yet powerful framework is proposed that is able to reconstruct both geometry and appearance of a given object interactively. Working in an incremental mode of operation, it enables the user to reconstruct a given scene with full visual feedback during the progress of the reconstruction process. For geometry reconstruction, an existing surface reconstruction algorithm has been investigated and adjusted to the needs of the framework. Furthermore, a hardware-accelerated surface light field algorithm has been integrated into the framework that performs appearance reconstruction of the object. The target application of our framework is the reconstruction of real-life objects using mobile acquisition devices. We demonstrate the performance and usefulness of our framework by reconstructing models from previously acquired datasets of real-life objects. Furthermore, we provide results of experiments run in our own model acquisition setup.