Deep learning, inspired by the human brain, is adept at deciphering extensive data sets. In recent years, its methodologies have gained prominence across diverse domains, notably in image processing. These methods shine in tasks such as object detection, image augmentation, and especially in super-resolution-a technique that transforms low-resolution images into high-definition versions. Prominent models like the Convolutional Neural Network for Super Resolution, Super Resolution Residual Network, Generative Adversarial Network for Super Resolution, and its enhanced version, are pivotal in this transformation process. Each layer in these models is tailored to refine low-quality images into their high-resolution analogues. This means multiple clear images can emerge from a single blurry counterpart. For our research, we compiled a dataset of assorted images. We then compared the original low-quality versions to their deep learning-enhanced equivalents. To ensure a rigorous comparison, we employed metrics like the Peak Signal-to-Noise Ratio and the Structural Similarity Index. Our evaluation also incorporated calculating the 'ground truth' value, ensuring a comprehensive assessment. Among the models tested, VGG54 was a standout, boasting an impressive 96% similarity index. This underscores deep learning's potential in revolutionizing image processing and enhancement.