Non Linear Tensor Diffusion Based Unsharp Masking for Filtering of COVID-19 CT Images

Abstract

COVID-19 is a dreadful disease caused by coronavirus and it belongs to the family of single-stranded RNA viruses. The Computed Tomography (CT) imaging was found to be a primary diagnostic tool in the screening of COVID-19. Preprocessing is the first stage in image processing operation, it improves segmentation and classification accuracy and hence it gains importance. Preprocessing techniques plays vital role in the improvement of image quality and the objective is to minimize noise, elimination of artifacts and aliasing effects. The improved contrast aids image segmentation and compression algorithms for better diagnosis by physicians. The CT images in general are corrupted by Gaussian and salt and pepper noise. The classical filtering techniques are median, Gaussian, bilateral and anisotropic diffusion. This chapter proposes a novel filtering technique, Non Linear Tensor Diffusion based Unsharp Masking for CT images. The performance validation was done by performance metrics like Just Noticeable Distortion (JND), Discrete Entropy (DE) and average mean brightness error (AMBE) for comparative analysis, classical filtering algorithms are used. The filtering algorithms are implemented in Matlab2015b and tested on real time CT images of COVID-19.