Blind deconvolution of blurred images with fuzzy size detection of point spread function

Abstract

The objective of present research is to find the correct dimensions of point spread function (PSF), which is a crucial step in blind deconvolution of linearly blurred images. Usually size of PSF is estimated by trial and error, based on user experience. Keeping in view the fuzzy nature of this problem, we have implemented a fuzzy inference system (FIS) in Matlab to efficiently predict the size of PSF. The fuzzy system is based on the common observation that the size of PSF is directly related to the amount of degradation caused to each pixel in the original image. The blurred image is compared with edge extracted image and the PSF size is estimated by accounting for the distortion of edges. The results are encouraging and the method presented in this paper can be used to avoid trial and error based lengthy process of PSF size estimation. © 2012 Springer-Verlag.