Selective noise filtration of image signals using wavelet transform

Abstract This paper examines the technique of denoising and signal extraction using wavelet transform scale space decomposition. The noisy signal is decomposed into multiple scales by the dyadic wavelet transform. At a given position, the level of correlation is used to deduce the presence or absence of significant feature of signals or images, which is then passed through the filter. By comparing the correlation information of the data at that scale with those at larger scales, noise is preferentially removed from the data. In the wavelet transform domain, the desired features are identified and retained because they are strongly correlated across scales compared to noise. This denoising algorithm can be used to reduce noise to a high degree of accuracy, while preserving most of the important features of the signal. In this paper, this technique of scale space filtering is applied to sample signals and images. Representative results are presented which shows that considerable noise content in signals and images can be reduced while preserving the value of the desired signal.