New images of a three-dimensional scene can be generated from known image sequences using lightfields. To get high quality images, it is important to have accurate information about the structure of the scene. In order to optimize this information, we define a residual-function. This function represents the difference between an image, rendered in a known view from neighboured images and the original image at the same position. In order to get optimal results, we minimize the residual-function by defining a nonlinear least-squares problem, which is solved by an appropriate optimization method. We use a nonmonotone variant of the Levenberg-Marquardt method.
The Preprint-Series of the Institute of Applied Mathematics, Preprint No. 322, 2007