Papyrus
http://hdl.handle.net/1866/16701
Abstract
One of the most visible aberrations in an image taken with a camera is radial distortion. When straight objects in the real world are projected onto an image as curved objects, this curvature is generally due to radial clistortion. Since computer vision uses these images for measurment, it is important to quantify this cleformation. This thesis concentrates on radial-clistortion calibration. Hence, the goal of the algorithm described is to measure the amount of radial distortion in one or many images acquired with the same lens. The method reported in this document differs from others in several ways. First, it is based on the radial distortion’s geometric properties. Second, the problem is solved with linear equations. Third, the center of distortion is cornputed, not fixed. finally, the solution isn’t model-based; conseciuerrtly, calibration of more general functions is possible.
Documents