A Hybrid Restoration Approach of Defocused Image Using MGAM and Inverse Filtering

A novel hybrid restoration scheme of defocused image is presented, which uses multivariate generalized additive model (MGAM) which is a nonparametric statistical regression model with no curse of dimensionality and inverse filtering (InvF). In this algorithm, firstly the five features of wavelet domain in defocused digital image, which are very stable relationship with the point spread function (PSF) parameter, are extracted by training and fitting a multivariate generalized additive model which is to estimate defocused blurred parameter. After the point spread function parameter is obtained, inverse filtering, which is needed to known the point spread function and a non-blind restoration method, is applied to complete the restoration for getting the true image. Simulated and real blurred images are experimentally illustrated to evaluate performances of the presented method. Results show that the proposed defocused image hybrid restoration technique is effective and robust. Index Terms — Defocused Image Restoration; Wavelet Transform; Multivariate Generalized Additive Model (MGAM); Inverse Filtering (InvF)