WSEAS Transactions on Systems
Print ISSN: 1109-2777, E-ISSN: 2224-2678
Volume 12, 2013
Wavelets and Ridgelets for Biomedical Image Denoising
Authors: , ,
Abstract: Image de-noising is a key step in the processing of medical images as they are often corrupted by noise in the process of receiving, coding and transmission. In this paper the performance of Discrete Wavelet Transform (DWT) (Bivariate shrinkage), Stationary Wavelet Transform (SWT) (hard thresholding), Dual Tree Complex Wavelet Transform (DTCWT) (Bivariate shrinkage) and Ridgelet Transform (Hard thresholding) for biomedical image de-noising are evaluated and compared in terms of Peak Signal to Noise Ratio (PSNR). The DWT in many applications reaches its limitations such as oscillations of coefficients at a singularity, lack of directional selectivity in higher dimensions, aliasing and consequent shift variance. Therefore SWT and DTCWT, both with their shift invariant property are studied. DTCWT a moderately redundant multi-resolution transform with decimated sub bands runs into two DWT trees (real and imaginary) of real filters producing the real and imaginary parts of the coefficients. A locally adaptive de-noising algorithm using the bivariate shrinkage function is illustrated using both DWT and DTCWT. A simple bivariate shrinkage rule is described to model the statistics of wavelet coefficients of images. The Ridgelet transform was developed over several years to break the limitations of Wavelet Transform and possesses high directional selectivity. Simulations and experimental results demonstrate that the DTCWT outperforms SWT and DWT as well as Ridgelets in denoising biomedical images corrupted by Random noise, Salt and pepper noise and Gaussian noise while SWT outperforms other wavelet techniques and Ridgelets in de-noising biomedical images degraded by Speckle noise and Poisson noise.