The Medical imaging processing has become fundamental in computer vision and often used in diagnosis of diseases the two are skin cancer and brain tumors. This technology has greatly contributed to limit the load on care institutions and staff since COVID-19 appeared. The identification of COVID-19 has its unique difficulties, especially as to its symptoms in patients with similar diseases such as flu. Therefore, this paper presents a new approach to categorize lung diseases. From the proposed (CNN) structure employed to conduct a deep analysis of chest X-rays, the differences of all lung diseases can be discerned easily. This technique is applicable to analyzing chest X-rays and differentiating features of the given condition in detail. The CNN model is applied to X-rays in order to identify the specific features of lung diseases versus COVID-19 with the main goal of developing an effective differential diagnosis for the diseases. The first applied operation was integer Sobel transform detects edges by demonstrating intensity gradients and contours the data was split into two sets: Specifically, the first part of the data was employed for building the model, the second one was used to validate this model. With K-Fold cross-validation the presented model was trained based on CNN architecture and then validate. Confusion matrix used to assess performance of several parameters with F-score, which combines precision and recall to provide a balanced assessment of model performance, overall accuracy was 94.39% indicating how useful this approach can be when it comes to distinguishing between X-rays COVID-19, X-rays normal and X-rays other lung diseases.