In this study, CNNs and auto encoders are combined to handle super resolution photos and videos by using their own unique benefits in extraction of features and representation learning. During developing a proposal, a bidirectional deep learning design that combines CNNs’ understanding-related skills with automatic encoders’ reconstruction abilities is being considered. Our algorithm learns complex spatial and structural properties by being trained on a dataset of matched high-resolution and low-resolution photos and videos. In the process of training, the model produces visually appealing high-resolution outputs while simultaneously improving the resolution and preserving important picture features. We demonstrate the efficacy of our technique on many benchmark datasets through thorough testing and assessment, displaying notable gains in visual quality, the edge maintenance, and fine sophisticated reconstruction. We also examine how the model may be adjusted to different super-resolution factors, which makes it a flexible tool with a wide range of real-world uses. By utilizing the complementary abilities of CNNs and auto encoders, this study advances picture Super-Resolution approaches and delivers a secure and efficient technique for improving the appearance and resolution. The suggested method has potential for use in a variety of fields, including electronic media, imaging in medicine, and surveillance.