A system study of sparse aperture sensors in remote sensing applications with explicit phase retrieval

Publication Year:
Usage 58
Downloads 46
Abstract Views 12
Repository URL:
Daniel, Brian
Dilute apertures; Hyperspectral; Image quality; Phase diversity; Phase retrieval; Sparse apertures
thesis / dissertation description
Sparse aperture designs use multiple primary mirrors in alignment to increase optical resolution, as opposed to using just one bigger primary mirror. Sparse apertures are defined to have a low amount of active optical area compared to the encircling area of the array. Previous works on sparse aperture systems have studied several configurations based on the gray-world assumption, where the spectral nature of the object, or scene, is assumed independent of its spatial location. In other words, the scene is the same color for every pixel. Robert Introne, Ph.D. (2004) and Noah Block, M.S. (2005) have tested this assumption and showed that in aberrated systems, or in extremely spectrally diverse scenes, the assumption fails. This work utilizes the spectral world model developed by Introne, but includes the Broadband Phase Diversity algorithm for phase retrieval. Phase retrieval mitigates misalignment errors between each of the sub apertures of the system that are extremely detrimental to the final image quality. Phase diversity uses multiple images of the same scene with the same system, except for one difference. Each image is captured with a different amount of a known additional aberration. Common choices are piston and defocus. This work will utilize defocus diversity. With the introduction of the defocus in the imagery, it is possible to estimate the misalignment errors (and other aberrations) of the optical system using minimization techniques. Knowledge of the misalignment errors aids in the restoration of the image. The gray-world assumption is the cornerstone of Broadband Phase Diversity (BPD). While the validity of the assumption has been examined, the effect of misalignment aberrations and spectrally diverse scenes on BPD performance has not been investigated until now. A trade-space is developed varying the aberration, the noise, and the bandwidths of imaging scenarios. Multiple scene types to simulate different levels of spatial and spectral variability were used to find image-independent trends in BPD performance and image quality. Human visual system based metrics were used to evaluate image quality. The relationship between BPD performance and image quality is examined. Methods of characterizing a hyperspectral image cube were developed in order to describe the spectral character that causes the gray-world assumption to fail. A model comparison was made between the spectral-world and gray-world forward models to evaluate the limits of the gray-world model.