Medical Image Denoising Based on Stationary Wavelet Transform and Soft Shrinkage Rule

Image denoising is one of the most important steps in image and video processing applications. Digital images are corrupted by various types of noise during acquisition and transmission. The Stationary Wavelet Transform (SWT) is an enhanced version of the discrete wavelet transform (DWT). SWT overcomes the lack of translation invariance present in DWT by removing the down-samplers and up-samplers. In this paper, an innovative approach for image denoising based on SWT is proposed where the effects of noise are minimized by soft thresholding on high frequency sub-bands. In this proposed methodology, scaled mean absolute difference (MAD) is calculated from which threshold value(T) is deduced for implementing the algorithm for minimization of noise effects. We exercised our methodology on four different medial images and obtained Peak Signal to Noise Ratio (PSNR) for various noise variances ranging from 10 to 30. Experimental results show that the proposed method gives significant Peak Signal to Noise Ratio (PSNR) values preserves the image edge information as well. It i also observed that the time taken for computation is almost same for all the images for each noise variance level.