Histogram-Based Gaussian Noise Removal

In modern science and technology, as the digital image processing gets more and more importance, the process of image quality enhancement and image restoration becomes a matter of concern for the researchers. In an image, denoising complexity increases from salt and pepper impulse noise to random-valued impulse noise, through to Gaussian noise. As salt and pepper noise occupies either high (255) or low (0) values and also as it is not distributed uniformly, hence, with the help of correct neighbouring pixels and known value of noise, the corrupted pixels can be easily restored. In random-valued noise, the value of noise is unknown, but we can take advantage of the non-uniform noise distribution, and using non-corrupted neighbouring pixels values, we can restore the corrupted pixels. However, as Gaussian noise is a uniform noise, hence, not only the value of noise is unknown, but also it is not possible to get non-corrupted neighbouring pixels. In Gaussian noise, restoring pixels using the corrupted neighbouring pixels is a highly tedious job. In this paper, an efficient algorithm for the removal of Gaussian noise is proposed. Experimental results show that proposed algorithm produces good results up to 50% of noise level, the value normally real-time images were corrupted.