A novel approach to blind deconvolution based on generalized Akaike’s information criterion

We propose a generalized version of Akaike's information criterion (AIC) as a novel criterion for estimating a point spread function (PSF) from the degraded image only. We first show that the generalized AIC (G-AIC) is equivalent to quadratic prediction loss up to some constant, and prove that incorporating exact smoother filtering, the minimization of the prediction loss yields exact estimate of PSF. The PSF is obtained by minimizing this G-AIC over a family of approximated smoother filterings. Based on this estimated blur kernel, we then perform non-blind deconvolution using our recently proposed SURE-LET algorithm. The proposed framework is exemplified with a number of parametric PSF. The experimental results demonstrate that the minimization of this criterion yields highly accurate estimates of the PSF parameters, which also result in a negligible loss of visual quality, compared to that obtained with the exact PSF. The highly competitive results show the great potential of developing more powerful blind deconvolution algorithms based on this criterion.