Stray Light Compensation in Optical Systems

Abstract: All optical equipment suffers from a phenomenon called stray light, which is defined as unwanted light in an optical system. Images contaminated by stray light tend to have lower contrast and reduced detail, which motivates the need for reducing it in many applications. This master thesis considers computational stray light compensation in digital cameras. In particular, the purpose is to reduce stray light in surveillance cameras developed by Axis Communications. We follow in the spirit of other digital stray light compensation approaches, in which measurements are fit to a parametric shift-variant point spread function (PSF) describing the stray light characteristics of the optical system. The observed contaminated image is modelled as an underlying ideal image convolved with the PSF. Once the PSF has been determined, a deconvolution is performed to obtain a restored image. We provide comparisons of a few deconvolution strategies and their performances regarding the restoration of images. Also, we discuss different techniques for decreasing the computational cost of the compensation. An experiment in which the images are compared to a ground-truth is proposed to objectively measure performance. The results indicate that the restored images are closer to the ground-truth compared to the observed image, which implies that the stray light compensation is successful.