%0 Generic
%A Geese, Marc
%D 2013
%F heidok:14391
%R 10.11588/heidok.00014391
%T Image Sensor Nonuniformity Correction by a Scene-Based Maximum Likelihood Approach
%U https://archiv.ub.uni-heidelberg.de/volltextserver/14391/
%X Image sensors come with a spatial inhomogeneity, known as Fixed Pattern Noise or image sensor nonuniformity, which degrades the image quality. These nonuniformities are regarded as the systematic errors of the image sensor, however, they change with the sensor temperature and with time. This makes laboratory calibrations unsatisfying. Scene based nonuniformity correction methods are therefore necessary to correct for these sensor errors.   In this thesis, a new maximum likelihood estimation method is developed that estimates a sensor’s nonuniformities from a given set of input images. The method follows a rigorous mathematical derivation that exploits the available sensor statistics and uses only well-motivated assumptions. While previous methods need to optimize a free parameter, the new method’s parameters are defined by the statistics of the input data. Furthermore, the new method reaches a better performance than the previous methods. Specialized developments that include a row- or column-wise and a combined estimation of the nonuniformity parameters are introduced as well and are of relevance for typical industrial applications.   Finally it is shown that the previous methods can be regarded as simplifications of the newly developed method. This deliberation gives a new view onto the problem of scene based nonuniformity estimation and allows to select the best method for a given application.