In this dissertation impulsive noise elimination methodologies are studied and four novel schemes are proposed. We have followed the two stage process of detection of noisy pixels followed by filtering of the same, in all our proposed schemes. Thus avoiding unnecessary filtration of non-noisy pixels. Threshold value is chosen adapting to the environment to make the detection process more efficient. Soft computing techniques are extensively used in the proposed techniques to make the threshold selection adaptive. Further we suggest an enhancement scheme under noisy conditions.