Analysis and Modeling of Passive Stereo and Time-of-Flight Imaging

This thesis is concerned with the analysis and modeling of effects which cause errors in passive stereo and Time-of-Flight imaging systems. The main topics are covered in four chapters: I commence with a treatment of a system combining Time-of-Flight imaging with passive stereo and show how commonly used fusion models relate to the measurements of the individual modalities. In addition, I present novel fusion techniques capable of improving the depth reconstruction over those obtained separately by either modality. Next, I present a pipeline and uncertainty analysis for the generation of large amounts of reference data for quantitative stereo evaluation. The resulting datasets not only contain reference geometry, but also per pixel measures of reference data uncertainty. The next two parts deal with individual effects observed: Time-of-Flight cameras suffer from range ambiguity if the scene extends beyond a certain distance. I show that it is possible to extend the valid range by changing design parameters of the underlying measurement system. Finally, I present methods that make it possible to amend model violation errors in stereo due to reflections. This is done by means of modeling a limited level of light transport and material properties in the scene