The DNN model is implemented for the automatic detection of various kinds of diseases and abnormalities in different image modalities. Automatic detection of COVID-19 based on DNN models are utilized in this work, along with the four distinct image processing models namely, convolution neural network, VGG-19, Inception, and Xception are applied on publicly available 402 COVID-19 infected and 397 normal images of lung CT-scan images. A total of eight different gradient-based features are also extracted to train thetraditional machine learning models. The Classification metrics of the deep learning model is also juxtaposed with machine learning models. The highest 93.887% and 94.092% training and testing accuracy has been achieved by applying the VGG-19 based deep learning model. The Image Processing models outperform the traditional DNN model. The research proposes a novel approach which is over-fitting independent and show accurate classification of covid-19 infection from normal diseases like Pneumonia using CT-Scan images of lungs.