New Physical Constraints for Multi-Frame Blind Deconvolution.

摘要

This work extended on previous efforts to explore state-of the-art blind image deconvolution techniques for improving space surveillance capabilities. The project had two objectives: (1) to develop and test a new theory of speckle statistics based on the model of partially-developed speckle, and (2) to test if inclusion of this theory in deconvolution removes the need for human intervention. Using a physics-based approach to exploit knowledge of the whole imaging process, the goal is to have a fully automatic process that could eventually approach real-time capabilities. The first year resulted in an Adaptive Wavelets Maximum Likelihood Estimator (AWMLE) algorithm for adaptive optics assisted observations (public release) while the second year focused on modelling of spatial, temporal and spectral integration of speckle statistics. Including constraints in Fourier domain, this work led to a novel methodology of testing the reconstruction fidelity in the image reconstruction process. Overall, this project has developed new models for statistics of speckle images based on a theory which is fundamentally different from the approaches developed in the 1970s. When checked against baseline theory, this new multi-frame blind deconvolution approach proved extraordinarily good performance while exhibiting autonomous capabilities.