Image Noise Level Estimation by Employing Chi-Square Distribution

Image denoising is a hot topic in computer vision. Most prior image denoising methods reduce noise with the empirical noise level, resulting in noise removal uncleanly or structure smoothness. The main reason is that the noise level is not precisely and adaptively estimated. To cope with this problem, this paper proposes a novel noise estimation scheme based on Chi-square distribution, including: First, the noisy image is divided into image blocks through a sliding window. Then, the flat blocks are selected from the image blocks by using the gradient feature. Next, the initial noise level is estimated by chi-square distribution on flat blocks. Finally, the stable noise level is final obtained by an iterative strategy. Experimental results validate that the proposed noise level estimation strategy is effective and is even superior to the state-of-the-arts.