Morphological Operations For denoising of White Gaussian Noise Corrupted MR Images
Varun Gandhi, Vansh Mendiratta, Sumedha Thakur, Mahipal Singh Choudhry · 2020 International Conference on Electronics and Sustainable Communication Systems (ICESC) · 2020
MRI scan is an extremely accurate method for pathological disease detection and it provides crucial information of tissues inside the human body. It is a painless technique that uses strong magnetic fields along with radio waves to generate precise images that provide accuracy in detecting structural abnormalities within the body. Very often, these MR images get corrupted by different noises of different probability density functions, and to get accurate information, denoising needs to be performed on these corrupted images. The most common types of noises associated with MR images are White Gaussian Noise, Salt and Pepper Noise and Speckle Noise. This paper discusses the denoising of MR images corrupted with White Gaussian Noise with a zero mean and uniform normal distribution, i.e, distributed independently and identically over an MR image. The denoising of corrupted grayscale MR images is carried out using five different morphological operations- Erosion, Dilation, Opening, Closing and Averaging alongside seven most commonly used structuring elements of size 3×3 and 5×5. After the analysis, it is observed that the Erosion and Opening operations are best suited for MR Images corrupted with white Gaussian noise as they give the most efficient results in terms of Signal-to-Noise ratio, Correlation coefficient, and computation time. Pertaining to structuring elements, 5×5 proves to be better at denoising the said Image than the commonly used 3×3 structuring elements.