An Adaptive Filter for Image Denoising using Fuzzy Inference

E.S. Hore, B. Qiu, and H.R. Wu (Australia)


Gaussian noise restoration, fuzzy inference system,weight median filter, L- lter


In this paper an image denoising algorithm is presented that is aimed at restoring images corrupted with Gaussian noise. A combination of a fuzzy inference system together with a number of statistical measures is used in the calculation of the nal output. The fuzzy inference system is employed to calculate a set of coefficients which are then used in a linear summation of a number of sub filters for the nal output calculation. The sub lters consist of two weighted median lters and two neighbours to the centre pixel in an ordered set. The ordering operation is carried out using the total inter-pixel differences as the measure rather than the raw pixel value. This allows the system to use the indices at which each pixel is placed in the new set as measures for determining the amount of noise present in the image pixel. As shown in the results the proposed scheme provides a promising framework for future work in extending the algorithm to vector valued images (i.e. colour images)

Important Links:

Go Back