| Popis: |
This dissertation is concerned with multiple-frame (multiframe) reconstructions using imagery acquired in dynamic imaging environments. Through several interesting examples, we address and relate the key concepts of information weighting, channel diversity and multiframe processing in the context of producing high resolution estimates from severely degraded imagery. For the problem of space object identification, we look at methods for preprocessing a collection of atmospheric turbulence-degraded short-exposure images to improve the resolving power of estimation algorithms. Specifically, we examine the performance of using frame selection to extract the least degraded subset of images from an ensemble for processing. Several measures of image quality are compared against idealized standards to demonstrate their relative effectiveness for ranking highly the least degraded image frames. We also examine the resolving implication of removing additive background noise, resulting from the sky and telescope. Specifically, we show that background compensation acts as a defacto restoration of the compact object support and leads to furthering the resolving power of estimation algorithms. In the context of dilute aperture imagery, we look at methods for inducing channel diversity into a collection of measurements. With a diverse image set, we compute estimates using both a joint multiframe objective and an aggregated objective. We then examine the implication of using joint or aggregate objectives in any estimation algorithm from a set-theoretic standpoint. Finally, we extend the classic Wiener filter for the multiframe case. The resulting formulation demonstrates that the appropriate weighting of image data allows for the worst frames to be included while improving the restoration. We discuss how this contradicts the earlier idea of frame selection and relates the multiframe Wiener filter to the dual information theoretic concept of "water-filling". |
