Mesostructure from specularity using a raster display and a digital camera

Abstract

This thesis describes a method for efficient acquisition of fine-scale surface details, or mesostructure surfaces, where a single camera is used to capture the mesostructure illuminated by a raster display. When adopting a typical camera-screen setup where a camera is placed on top of a raster display, a camera calibration step is inevitable. Camera calibration allows mapping from camera pixel coordinates to rays in the scene, and from points in the scene to camera pixel coordinates. To establish the relation between the screen and the camera, different screen calibration methods are discussed, using a planar or spherical mirror. A concise overview of popular shape recovery techniques is presented, including shape from shading, photometric stereo and structured light. Because these techniques are usually inadequate to reconstruct highly specular fine-scale details, shape reconstruction using specularities is examined. Starting from the general shape reconstruction from specularities the more specialized case of mesostructure from specularity is discussed. Mesostructure from specularity requires a high directional light resolution resulting in a high acquisition time. We propose a new multiplexed illumination scheme using gray code patterns. This technique exploits specularity properties, allowing for a fast and inexpensive shape recovery. The acquisition of N light sources is reduced to O (log2 N). A raster display functions as a controllable planar illuminant which makes it easy to emit patterned illumination. Further the implementation of the discussed methods is presented. Results obtained from our proposed multiplexed illumination scheme are thoroughly discussed, presenting both synthetic and real-world examples.