Blind Image Restoration Based on Signal-to-Noise Ratio and Gaussian Point Spread Function Estimation

Abstract

In order to improve the quality of restored image, a blind image restoration algorithm is proposed, in which both the Signal-to-Noise Ratio (SNR) and the Gaussian Point Spread Function (PSF) of the degraded image are estimated. Firstly, the SNR of the degraded image is estimated through local deviation method. Secondly, the PSF of the degraded image is estimated through error-parameter method. Thirdly, Utilizing the estimated SNR and PSF, high resolution image is restored through Wiener filtering restoration algorithm. Experimental results show that the quality and peak signal-to-noise of the restored image are better around the real value and justify the fact that the SNR an-d PSF estimation plays great important part in blind image restoration.

A New Method for Ultrasound Detection of Interfacial Position in Gas-Liquid Two-Phase Flow

“…Unfortunately, blind SNR estimation is an open issue in image processing theory. The ratio between maximum and minimum local variances as described in [17] can be used as a rough SNR estimate and it is defined as $\text{SNR}=10{log}_{10}\left(\frac{max\left({\delta }_I^2\right)}{min\left({\delta }_I^2\right)}\right)$where ${\delta }_I^2$ is the local variance of image I at position ( i , j ) defined as ${{\delta }_I}^2\left(i,j\right)=\frac{1}{\left(2p+1\right)\left(2q+1\right)}\sum _{k=-p}^p\sum _{l=-q}^q{\left[I\left(i+k,j+l\right)-{\mathrm{normal\mu }}_I\left(i,j\right)\right]}^2$where p and q are the sizes of the local area, I is an image formed by the spatiotemporal distribution of the considered shape parameter and μ I the local mean value which is defined as ${\mathrm{\mu }}_I=\frac{1}{\left(2p+1\right)\left(2q+1\right)}\sum _{k=-p}^p\sum _{l=-q}^{q}I\left(i+k,j+l\right)$in this study the local area size was defined as p = 2 and q = 2, as suggested in [17]. …”