In the event of the COVID-19 outbreak, prompt and accurate diagnosis became critical for both public health interventions and efficient patient care. COVID-19 is a disease that affects the upper and lower respiratory tract and can have fatal consequences. Early diagnosis is crucial for effective treatment and containment. Studies have shown that COVID-19 manifests in the chest of infected patients, prompting the computer vision community to explore the use of CT scans and deep learning-based solutions for diagnosis. However, efforts to implement explainable artificial intelligence (AI) for interpreting deep learning models in COVID-19 recognition are still scarce. In this paper, we apply SHAP (Shapley Additive Explanations) and LIME (Local Interpretable Model- agnostic Explanations) techniques to enhance the interpretability of our developed CNN that used to detect COVID-19 from CT scan images. The dataset is consist of 4649 images, of which 2476 are from patients with COVID-19 and 2173 are from patients without COVID-19 was used in our implementation. We applied SHAP and LIME techniques to identify important features of COVID-19 images, which even improved the performance of our original model. The comparison results with other baseline models show the robustness of our proposed model and identified important features. We also find that the explainable ability of the SHAP and LIME techniques also depends on its model prediction accuracy.