Review of Recent Developments in Image Blind Deconvolutio n
Jan Kotera · 2012
Blind deconvolution (or deblurring) attempts to recover the original sharp photograph from a blurred one. Such task is very underdetermined and is therefore usually solved as maximum a posteriori probability (MAP) estimation, incorporating prior information on natural images. We provide a detailed intro- duction to the subject, formulate blind deconvolution as MAP estimation, and describe how individual components appeared in recent literature. We pay special attention to the distinction between traditional simultaneous estimation of blur kernel and sharp image as opposed to estimation of the kernel alone by taking into account all possible sharp images. These two branches emerged in recent papers and understanding their relationship seems to be the way towards successful deconvolution.