Computational imaging: The improved and the impossible

While the performance of optical imaging systems is fundamentally limited by diffraction, the design and manufacture of practical systems is intricately associated with the control of optical aberrations. The fundamental Shannon limit for the number of resolvable pixels by an optical aperture is generally therefore not achieved due to the presence of off-Axis aberrations or large detector pixels. We report how co-called computational-imaging (CI) techniques can enable an increase in imaging performance using more compact optical systems than are achievable with traditional optical design. We report how discontinuous lens elements, either near the pupil or close to the detector, yield complex and spatially variant PSFs that nevertheless provide enhanced transmission of information via the detector to enable imaging systems that are many times shorter and lighter than equivalent traditional imaging systems. Computational imaging has been made possible and attractive with the trend for advanced manufacturing of aspheric, asymmetric lens shapes at lower cost and by the exploitation of low-cost, high-performance digital computation. The continuation of these trends will continue to increase the importance of computational imaging.