Development of Noise Suppression Schemes in Images

Abstract

Noise suppression from images is one of the most important concerns in digital image processing. Two important noise models are considered in this thesis i.e. random valued impulse noise and Gaussian noise and two propositions have been made to suppress these noises. The first scheme is detection based filtering which uses the Bayes classification technique to detect the noisy pixels. The detected noisy pixels are then filtered out using a weighted median filtering. In another scheme an attempt has been made to improve the existing spatially adaptive denoising algorithm for suppression of Gaussian noise. The proposed scheme uses uniform weighting coefficients and utilizes local statistics parameters to detect as well as to filter the noisy pixels. The suggested scheme gives good results for high level Gaussian noise. Extensive simulations on standard images are carried out to show the efficiency of the proposed schemes along with other state of the art techniques under similar environment. Subjective as well as objective performance comparisons show the better noise suppression capability of the proposed algorithms than their counterparts.