In the modern world, lung diseases are most common due to various factors like environmental issues, pandemics etc. and these diseases can be mild (influenza) or severe (lung cancer) etc. for the identification of lung disorders it is necessary to use imaging modalities like chest radiographs (CXRs), CT and MRI so that the infected area can be easily identified. In this paper, a computationally efficient TL-based method is used to segment lung regions from CXRs. The proposed method utilizes U-Net-based encoder-decoder architecture along with the TL mechanism having EfficientNetB7 as encoder which is pre-trained on ImageNet. The method’s effectiveness is assessed using several quantitative metrics and has obtained the highest accuracy and F1-Score of 98 and IoU of 96 on the JSRT dataset. The performance is further matched with state-of-art CNN architecture and obtained better results in Accuracy and IoU parameters.