Impulse noise reduction based on lifting wavelet transform

This paper presents a novel method to restore signal corrupted with impulse noise by using time-varying lifting wavelet filters (TVLWF). It is difficult to remove such a noise by conventional linear niters and the Donoho's wavelet shrinkage method. Using the TVLWF, we can detect the locations of impulse noise and estimate their amplitude by which impulse noise can be removed. Proposed noise reduction method is tested on an electrocardiogram (ECG) signal corrupted with mixed Gaussian and random-valued impulse noise. Major improvement is obtained in comparison with conventional median filter.