Author(s): Sarah B. Aziz and Maytham A. Shahed
Article publication date: 2009-09-01
Vol. 27 No. 3 (yearly), pp. 156-167.
166
Keywords
Impulsive noise, Median filter, Average filter, Mamdani neurofuzzy, Adaptive backpropagation
algorithm.
Abstract
The results of the noise removal have a strong influence on the quality of the other
image processing techniques. And since the main goal for any noise removal method is the
preservation of the edges and image information, this influence is the principal drawback of
the smoothing filters. In this paper, a modified Mamdani neurofuzzy network scheme has been
proposed and presented to improve effects of smoothing filters in removing impulsive noise. The
number of connections is reduced to be equal to the number of membership functions plus one.
A training strategy for the presented neurofuzzy is based on artificial image which reduces the
time of computation. The Analysis shows that the proposed neurofuzzy works well in suppressing
impulsive at different noise ratios for grayscale and truecolor images.