Efficient cepstrum analysis based UNLM PSF estimation in single blurred image

Abstract

We propose a new Uniform Non-linear Motion (UNLM) Point Spread Function (PSF) estimation algorithm based on cepstrum analysis. Our work has two contributions. First, the algorithm does not need an exhaustive selection of candidate PSFs as conventional algorithm does, only one PSF and its symmetric shape are evaluated for final decision. Second, we give theoretical reasoning, which was not clearly interpreted so far, on how to extent conventional Uniform Linear Motion (ULM) PSF algorithm to UNLM PSF estimation. We show the effectiveness of the proposed algorithm by both simulation and real images. Quantitative accuracy improvement to related work is also presented in the experiments.