Estimation of Noise Levels and Denoising of Mixed Noise in MRI Images Using Threshold Based DWT
Noise is an unwanted variation in the pixel values of an image. An important image preprocessing step is image
denoising, which removes noises from the image. Although many image denoising methods have been developed, denoising
remains a challenge because it results in side effects like artifacts and blurring the images. Magnetic resonance imaging
(MRI) is widely accepted imaging technique for plenty of medical applications. MRI is corrupted by Rician noise, and
Rician-Gaussian Noise mixture. In this paper a technique for estimating the noise level called Modified linear minimum
mean square error (MLMMSE) is proposed. In order to remove the noise from MRI images. A combination of threshold
based DWT with Modified K-Singular Value Decomposition (K-SVD) algorithm is proposed. DWT based denoising
technique is best suited for MRI Images, because of energy compaction, sparsity, and multi resolution. To enhance the result,
DWT is combined with improved K-SVD algorithm. In this work, for further denoising, along with existing thresholding
techniques such as Visushrink, Neighshrink, Sureshirnk, one more modified technique called Neighsureshrink is also
proposed. From the experimental results, it is observed that Neighsureshrink thresholding technique based Coif5 DWT in
association with improved K-SVD performs better.
Keywords- Medical Image Denoising, DWT, Improved K-SVD, Thresholding.