Noise estimation is fundamental and essential in a wide variety of computer vision, image, and video processing\napplications. It provides an adaptive mechanism for many restoration algorithms instead of using fixed values for the\nsetting of noise levels. This paper proposes a new superpixel-based framework associated with statistical analysis for\nestimating the variance of additive Gaussian noise in digital images. The proposed approach consists of three major\nphases: superpixel classification, local variance computation, and statistical determination. The normalized cut algorithm is\nfirst adopted to effectively divide the image into a set of superpixel regions, from which the noise variance is computed\nand estimated. Subsequently, the Jarqueââ?¬â??Bera test is used to exclude regions that are not normally distributed. The\nsmallest standard deviation in the remaining regions is finally selected as the estimation result. A wide variety of noisy\nimages with various scenarios were used to evaluate this new noise estimation algorithm. Experimental results\nindicated that the proposed framework provides accurate estimations across various noise levels. Comparing\nwith many state-of-the-art methods, our algorithm strikes a good compromise between low-level and high-level noise\nestimations. It is suggested that the proposed method is of potential in many computer vision, image, and video\nprocessing applications that require automation.
Loading....