Image Enhancement Methods and Applications in Computational Photography

• Main Author: Qian, Qinchun
• Other Authors: Gunturk, Bahadir
• Format: Others
• Language: en
• Published: LSU 2014
• Subjects: Electrical & Computer Engineering
• Online Access: http://etd.lsu.edu/docs/available/etd-12202013-220625/

Computational photography is currently a rapidly developing and cutting-edge topic in applied optics, image sensors and image processing fields to go beyond the limitations of traditional photography. The innovations of computational photography allow the photographer not only merely to take an image, but also, more importantly, to perform computations on the captured image data. Good examples of these innovations include high dynamic range imaging, focus stacking, super-resolution, motion deblurring and so on. Although extensive work has been done to explore image enhancement techniques in each subfield of computational photography, attention has seldom been given to study of the image enhancement technique of simultaneously extending depth of field and dynamic range of a scene. In my dissertation, I present an algorithm which combines focus stacking and high dynamic range (HDR) imaging in order to produce an image with both extended depth of field (DOF) and dynamic range than any of the input images. In this dissertation, I also investigate super-resolution image restoration from multiple images, which are possibly degraded by large motion blur. The proposed algorithm combines the super-resolution problem and blind image deblurring problem in a unified framework. The blur kernel for each input image is separately estimated. I also do not make any restrictions on the motion fields among images; that is, I estimate dense motion field without simplifications such as parametric motion. While the proposed super-resolution method uses multiple images to enhance spatial resolution from multiple regular images, single image super-resolution is related to techniques of denoising or removing blur from one single captured image. In my dissertation, space-varying point spread function (PSF) estimation and image deblurring for single image is also investigated. Regarding the PSF estimation, I do not make any restrictions on the type of blur or how the blur varies spatially. Once the space-varying PSF is estimated, space-varying image deblurring is performed, which produces good results even for regions where it is not clear what the correct PSF is at first. I also bring image enhancement applications to both personal computer (PC) and Android platform as computational photography applications.