Blind Image Restoration with Defocus Blur by Estimating Point Spread Function in Frequency Domain

A blind image restoration method is proposed to enhance the quality of defocus blurred image. Firstly, The defocused blurred image is converted to the frequency domain in order to analyze its spectral characteristics. Secondly, the defocus radius of the defocus point spread function is automatically estimated by calculating the distance between the center point and the nearest black circle. Thirdly, utilizing the estimated defocus point spread function, the image is blindly restored by the constrained least squares algorithm. Forthly, the impact of defocus point spread fuction on the restored image is tested. Experimental results show that the proposed defocus point spread function estimation algorithm has high accuracy, and the restored image achieves better visual effect and higher PSNR around the real defocus point spread function.