Abstract
In the field of medical image analysis, there is a substantial need forhigh-resolution (HR) images to improve diagnostic accuracy. However, It is achallenging task to obtain HR medical images, as it requires advancedinstruments and significant time. Deep learning-based super-resolution methodscan help to improve the resolution and perceptual quality of low-resolution(LR) medical images. Recently, Generative Adversarial Network (GAN) basedmethods have shown remarkable performance among deep learning-basedsuper-resolution methods. Real-Enhanced Super-Resolution Generative AdversarialNetwork (Real-ESRGAN) is a practical model for recovering HR images fromreal-world LR images. In our proposed approach, we use transfer learningtechnique and fine-tune the pre-trained Real-ESRGAN model using medical imagedatasets. This technique helps in improving the performance of the model. Thefocus of this paper is on enhancing the resolution and perceptual quality ofchest X-ray and retinal images. We use the Tuberculosis chest X-ray (Shenzhen)dataset and the STARE dataset of retinal images for fine-tuning the model. Theproposed model achieves superior perceptual quality compared to the Real-ESRGANmodel, effectively preserving fine details and generating images with morerealistic textures.