Effects of Photon Noise on Unconstrained Minimization Techniques for IterativeBlind Deconvolution

摘要

In recent years, imaging through atmospheric turbulence has interested militaryscientists seeking to improve optical surveillance of satellites. Adaptive optics was a step toward achieving diffraction-limited resolution from ground-based telescopes. Unfortunately, adaptive optics only partially compensate for atmospheric blurring, therefore post processing of images is required. Processing methods in use today require knowledge of the impulse response of the optical system to reconstruct imagery, but this information is seldom known. This thesis looks at a new metal of processing compensated imagery, called blind deconvolution, which assumes very little or no prior information about the impulse response. In particular, this investigation analyzes the performance of Lane's unconstrained minimization method of blind deconvolution. This technique is applied to simulated single and binary star images corrupted by photon noise. Results reveal that prior knowledge of the cutoff frequency of the system greatly enhances the ability of the algorithm to achieve accurate estimates of the object when measurements contain relatively few photo events. Additionally, this study discovered that estimates are highly dependent upon the choice of the support region. Analysis also shows that the algorithm produces estimates containing frequency content above the diffraction limit which may invalidate this method as a useful means to reconstruct imagery.