There has been a deluge of data-driven deep learning approaches to detect COVID-19 from computed tomography (CT) images over the pandemic, most of which use ad-hoc deep learning black boxes of little to no relevance to the actual process clinicians use and hence have not seen translation to real-life practical settings. Radiologists use a clinically established process of estimating the percentage of the affected area of the lung to grade the severity of infection out of a score of 0-25 from lung CT scans. Hence any computer-automated process that has aspirations of being adopted in the clinic to alleviate the workload of radiologists while being trustworthy and safe, needs to follow this clearly defined clinical process religiously. Keeping this in mind, we propose a simple yet effective methodology that uses explainable mechanistic modelling using classical image processing and pattern recognition techniques. The proposed pipeline has no learning element and hence is fast. It mimics the clinical process and hence is transparent. We collaborate with an experienced radiologist to enhance an existing benchmark COVID-19 lung CT dataset by adding the grading labels, which is another contribution of this paper, along with the methodology which has a higher potential of becoming a clinical decision support system (CDSS) due to its rapid and explainable nature. The radiologist gradations and the code is available at https://github.com/Samiran-Dey/explainable_seg .