Spatial Resolution Enhancement Technology Of Flat Compound-eye Imaging Systems

The flat compound-eye imaging system achieves thinness and multi-spectral, three-dimensional, digital focusing imaging by replacing a conventional full-aperture lens with a lenslet array. The reduction in the focal length and diameter of the lenslet(keeping the F number the same) leads to the poor performance in spatial resolution. However, the lenslet array could accumulate a series of subimages from the same scene, and these subimages can be processed by reconstruction algorithms to enhance spatial resolution. This thesis focuses on the spatial resolution enhancement of flat compound-eye imaging systems, a series of algorithmic and experimental researches about the technique are carried on. The main contents include the following:Firstly, with respect to the reconstruction algorithm, the following methods are proposed:(1) A Point Spread Function estimation algorithm based on USAF 1951 resoluion target is proposed, and the experiment demonstrates the feasibility.(2) A image registration agorithm, combined SIFT-RANSAC with adaptive support weight appoach, is presentd with higher sub-pixel displacement accuracy.(3) The thesis propoeses a acceleration strategy based on structural characteristics of the flat compound-eye imaging system. and the experiment validates its effectiveness.(4) A joint Maximum A Posteriori algorithm based on SSIM is proposed with better reconstruction performance.Secondly, the influence of the number of lenslets on the performance of reconstruction algorithms for the flat compound-eye imaging system is presented, and the lenslet number is optimized to achieve thin system and high spatial resolution performance. Subimages with different numbers of lenslets are generated following the observation model, and reconstruction algorithms are applied to evaluate the spatial resolution of the compound-eye system. The optimal lenslet number is determined via theoretical performance optimization and verified via experimental comparisons of spatial resolutions of three compound-eye systems. It is worth mentioning that registration errors always exist in registation algorithms. Therefore, the influence is analysed with actual translation parameters as well as translation parameters estimated by registration algorithms.Finally, The thesis focuses on experimental evaluation on spatial resolution enhancement technology. The spatial resolution with different F number are measured with the prototype of flat compound-eye imaging system and USAF 1951 resolution target of 3m object distance. Moreover, we present spatial resolution comparisons among the prototype and Nikon lenses in the same F number, same image sensor condition.