S. Manikandan  , D. Ebenezer  , 
    A New Detail Preserving Neural Median Filter with Reduced Streaking and False Alarm for Impulse Noise Suppression in Images,
    International Journal of Soft Computing,
    Volume 2,Issue 1,
    2007,
    Pages 186-190,
    ISSN 1816-9503,
    ijscomp.2007.186.190,
    (https://makhillpublications.co/view-article.php?doi=ijscomp.2007.186.190)
    Abstract: A new impulse detection algorithm is proposed which can successfully remove impulse noise from corrupted images while preserving image details; streaking is also reduced. The   algorithm is an appropriate combination of impulse detection algorithm and a Recursive Weighted Median [RWM] filter whose weights are calculated by back propagation algorithm based on universal approximation theorem. The main advantage of the algorithm is that it detects both positive and negative impulses with high accuracy while reducing the probability of falsely detecting image details as impulses. Results show that the proposed algorithm is more effective in terms of visual appreciation, noise suppression and mean squared error performance compared with standard median filtering, general recursive weighted median filtering and adaptive length median filtering.
    Keywords: Recursive weighted median filter;back propagation algorithm;universal approximation theorem;impulsive noise;adaptive window;noise removal