Suppression of impulse noise in MR images using artificial intelligent based neuro-fuzzy adaptive median filter

This paper presents a new artificial intelligent based neuro-fuzzy rule base adaptive median filter for removing highly impulse noise. Since the filter is rule base, it is called neuro-fuzzy rule base adaptive median (NFRBAM) filter. The NFRBAM filter is an improved version of switch mode fuzzy adaptive median filter (SMFAMF) and is presented for the purpose of noise reduction of images corrupted with additive impulse noise. The NFRBAM filter consists of a decision unit and three different types of filters. In the decision unit, the noisy input image is directed to the proper filter with respect to the noise density. Neuro-fuzzy rule based approach is used in both decision and filtering parts. In artificial neural network, multi layer perceptron (MLP) architecture with backpropagation (BP) algorithm is used for noise detection and removing highly impulse noise corrupted MR images. In fuzzy logic, bell-shaped membership function is employed in order to obtain better results. Experimental results indicate that the proposed filter is improvable with the increased fuzzy rules to reduce more noise corrupted images and preserve image details more than SMFAMF.