Blind image restoration (BIR) is a hot subject in image processing but hard to be applied due to its slow speed. To achieve real-time BIR, a new concept called the “mid-frequency region (MR)” of an image was proposed, which can be used to estimate the point spread function (PSF) and restore the image rapidly and precisely. In order to locate the MR effectively, a geometrical method without any parameters and special smoothing methods were proposed. And new MR-based Wiener filtering was derived to accomplish BIR quickly, too. Then, using the class G PSFs which fit numerous imaging systems, a new algorithm for real-time BIR was derived with the MR-based method and single-parameter derivative fitting, which was also accelerated with a fast method of smoothing. Experiments prove that the algorithm is speedy, stable, few-parameter and multifunctional, and can accomplish image deblurring, detail enhancement and noise suppression for real-time images.
