Computer Science · Computer Vision and Pattern Recognition
Image Restoration using Deep Learning
Dr. Hansaraj Wankhede, Abhishikth Thul, Malviya, Deep, Pratik Pathe, Shivam Kuite
See on the mapAbstract
Image restoration using deep learning<br> attempts to create an image recovery system that can<br> restore outdated and corrupted images regardless of the<br> type of noise present. Photos play an important role in<br> capturing and preserving cherished moments in today's<br> digital age. However, due to a variety of environmental<br> conditions, these images can get distorted over time.<br> Manual recovery is time-consuming and labor-intensive,<br> emphasizing the need for an automated alternative. Deep<br> learning techniques, notably convolutional neural<br> networks, are used in the proposed system, which has<br> demonstrated promising results in image processing<br> tasks. The essay goes over these approaches in detail,<br> focusing on image noise reduction, deblurring, dehazing,<br> and super-resolution. Different network topologies are<br> investigated, including those with residual or merge-skip<br> connections, as well as their receptive fields and the<br> usage of unsupervised autoencoder processes. The study<br> also looks at image quality criteria to see how helpful<br> they are in image recovery. An effective deblurring<br> network and adaptable training algorithms for highresolution recovery tasks are suggested to handle the<br> special difficulty of deblurring. The proposed method's<br> performance is compared to state-of-the-art approaches<br> utilizing both quantitative and qualitative assessments.<br> The study finishes with a discussion of prospective future<br> research topics and outstanding challenges in image<br> recovery. The ultimate goal of this research is to create a<br> robust and efficient image recovery system capable of<br> restoring photos to their original quality regardless of<br> the type or severity of corruption or noise present.