© 2021 IEEE.One of the primary methods to diagnose Covid-19 illness is to examine Chest X-ray images. In most patients, these images contain abnormalities caused by Covid-19 viral pneumonia. In this study, we conducted extensive empirical analysis to detect such pneumonia on images using Convolutional Neural Networks. Our analysis on a set of existing CNN models show that some of these models are insufficient in decision making. In this context, various binary class classification models are trained using the Covid-19 data from COVID-CXNet dataset and Normal class data from the NIH Chest X-ray dataset. We used Contrast Limited Adaptive Histogram Equalization (CHALE) and Bi-Histogram Equalization (BEASF) based on Adaptive Sigmoid Function for preprocessing the data. Using the transfer learning techniques, ImageNet pretrained models of various CNN models, i.e. DenseNet, VGGNet, EfficientNet are adapted to this domain. The best result is obtained, with a 0.99 F1 score, using the EfficientNetB5 model with preprocessed data.