Abstract: This paper uses neighborhood information of image pixels to estimate noise variance in a Gaussian white noise image. As there is a certain correlation among neighborhood pixels, it defines an algorithm about block neighborhood relevance in a blocking image. Using the neighborhood relevance it can select the smoothing regions in an image and estimate noise variance with the selecting smoothing regions. Experimental data presents that it is an effective method to estimate image noise.

Key words: image processing, block neighborhood relevance, image noise estimation, noise variance estimation