An active measurement, that senses the scene with a camera under controlled illumination, is one of the common techniques in computer vision field. Most of the conventional techniques rely on ideal light transport, or a response to the illumination of the scene. However, since an actual light transport depends on scene geometry and light behavior, undesired light transport such like inter-reflections and scattering could degrade the measurement.
In this thesis, we propose the acquisition of desired light transport by considering light paths. We specify paths by considering the geometry and the measurement setup, and selectively acquire the light transport along the path. Our contribution also includes analysis and application of the acquired light transport.
We tackle two problem settings. The first problem is an optical tomography of diffuse surface object that has a diffuse surface and a transparent body, like grapes. We propose acquisition of the light transport along straight paths inside the object. We propose a light path model and reconstruction algorithm. We demonstrate that our framework successfully reconstructs the interior of a real object. The second problem is a light transport acquisition of projector-camera system. We propose novel acquisition according to the transport distance. We demonstrate our acquired light transport can be applied for various visualization and application.