Variance-based no-reference quality assessment of AWGN images

In this paper, a no-reference quality assessment method for image contaminated with additive white Gaussian noise (AWGN) is proposed. The proposed spatial domain method is based on the fact that if the portion of an image having structure is distorted by AWGN, then variance of the distorted image increases. On the other hand, if the smooth portion of image is contaminated by AWGN, then the distorted image will have proportionate variance as that of AWGN. Therefore, variance of the smoother parts of the image contaminated with AWGN reflects the level of noise and hence can be used as an indicator of the image quality. Extensive simulation results show that the proposed method is highly accurate and lower in complexity in comparison to the existing algorithms.