A Level-Wavelet-Dependent Scheme for Image Denoising Via Undecimated Wavelet Transform

A.K. Ng, T.S. Koh, and C.H. Thng (Singapore)


Wavelet transforms, image denoising, and magnetic resonance imaging.


Denoising is a fundamental work in all image processing. This paper proposes a level-wavelet-dependent (LWD) scheme, which utilizes undecimated discrete wavelet transform and a hard-thresholding rule with a modified LWD threshold. Performance of the LWD denoising scheme are comparatively evaluated with the classical Donoho denoising scheme. Experiments conducted on both non-clinical and clinical images show that the LWD scheme could perform better than the Donoho scheme, yielding improved visual quality, lower mean-square error, and hence, higher peak-signal-to-noise ratio.

Important Links:

Go Back