Image quality enhancement using original lens via optical computing

High-end lenses are usually composed of many optical elements to compensate various optical aberrations, e.g. geometric distortion, monochromatic and chromatic aberrations. The resulting complexity and machining accuracy requirements make high-end lenses too expensive, heavy, and fragile for day-to-day photography. To address this problem, we devised an optical computing approach to touch-up the low quality photos produced by simpler lenses. We propose a setup consisting of an easily accessible display and the original camera in order to perform optical aberration correction with a deconvolution framework. The equivalence of the degeneration model and the lens's optical computing turns the traditional blind deconvolution algorithm into its non-blind counterpart and promises robust performance. A prototype system is implemented to verify the feasibility of the proposed method, and a series of experiments on both synthetic and captured images are applied to demonstrate effectiveness and performance.