A Computational Photography Algorithm For Quality Enhancement Of Single Lens Imaging Deblurring

Over the past decades, the requirement of high-quality image is increasing along with the development of technology, and the design of modern imaging optics has becoming increasingly complex. The lens of modern single lens reflex(SLR) cameras may contain a dozen or more individual lens elements, which are used to optimize light efficiency of the optical system while minimizing image blur caused by lens aberrations. The precision of lens design is getting more and more complex, despite their better geometric imaging properties than earlier lenses, modern lenses design face high cost and weight problems.Computational photography has revolutionized photography in recent years by presenting an integrative technology that combines computer and software technologies with modern sensors and optics to create new imaging systems and image applications. Inspired by this idea, this paper proposes an alternative way of capturing high-quality photographs without complex combined lenses. Only one simple single lens(plano-convex lens) is used for imaging and then the ensuing aberrations are corrected computationally, thus high quality image can also be achieved through single lens camera.This paper has proposed a computational photography algorithm for high quality single lens imaging. The research includes the following aspects:1. First a single lens imaging system is designed, then lens imaging principle is analyzed to understand the process of image blur caused by lens aberrations, thus the aberrations correction is transformed to the problem of image deblurring.2. A kernel structure prior has been proposed by analyzing the characteristic of single lens imaging system's point spread function. Along with it, an image color prior has also been proposed by analyzing the trait of image color changing.3. Combining the kernel structure prior and image color prior, a blind image deblurring algorithm based on Maximum a Posterior has been proposed to estimate the accurate point spread function, and then to improve the final quality of image deblurring.4. A fast point spread function calibration algorithm has been proposed, which make it possible for the application of single lens imaging system in normal life.Single lens imaging is quite a novel research direction in the field of both image processing and optics design, which is meaningful for research and important for normal life application. Our work is only the tentative exploration of this novel filed, more creative and painstaking work is needed to achieve the final mature applied technology.