A study on regression spline based local minima approach for gaussian noise reduction in images

Abstract

The study proposes a novel image denoising algorithm based on the regression splines (RS) for the restoration of images corrupted with the Gaussian noise. In the proposed algorithm, overlapping window of dimension 5×5 have been considered to replace the central pixel value with the local minimum of both diagonal pixels and central row and column pixels of the processing window. Selection of minimum of approximate pixel value helps in reducing the noise diffusion to the neighboring pixels. The proposed algorithm has been found to function efficiently for the Gaussian noise removal while preserving the fine image details.