Year of Publication


Document Type





Computer Science

First Advisor

Christopher O. Jaynes


This thesis characterizes the problem of relative camera calibration in the context of three-dimensional volumetric reconstruction. The general effects of camera calibration errors on different parameters of the projection matrix are well understood. In addition, calibration error and Euclidean world errors for a single camera can be related via the inverse perspective projection. However, there has been little analysis of camera calibration for a large number of views and how those errors directly influence the accuracy of recovered three-dimensional models. A specific analysis of how camera calibration error is propagated to reconstruction errors using traditional voxel coloring algorithms is discussed. A review of the Voxel coloring algorithm is included and the general methods applied in the coloring algorithm are related to camera error. In addition, a specific, but common, experimental setup used to acquire real-world objects through voxel coloring is introduced. Methods for relative calibration for this specific setup are discussed as well as a method to measure calibration error. An analysis of effect of these errors on voxel coloring is presented, as well as a discussion concerning the effects of the resulting world-space error.