In image processing, image super-resolution is an important process for redeeming high-resolution images from low-resolution images. Image super-resolution in computer vision has a wide range of real-life applications, such as medical imaging, text image enhancement, biometrics, remote sensing, media, satellite imaging, surveillance and security, and astronomical imaging.
Deep learning-based image super resolution models have been highly explored with the expansion of deep learning techniques in the modern years. Various deep learning methods have been employed to resolve image super-resolution tasks, spanning from the Convolutional Neural Networks (CNN) based method to Generative Adversarial Networks. Deep learning has an automatic feature extraction ability that significantly helps to enhance the Super-Resolution technology.
Supervised image super-resolution network designs are Residual Learning, Recursive Learning, Multi-path Learning, Dense Connections, Attention Mechanism, Advanced Convolution, Region-recursive Learning, Pyramid Pooling, Wavelet Transformation, xUnit, and Desubpixel. Loss Functions, Batch Normalization, Curriculum Learning, and Multi-supervision have supervised image super-resolution learning strategies. Zero-shot Super-resolution, Weakly-supervised Super-resolution, Learned Degradation, Cycle-in-cycle Super-resolution, and Deep Image Prior are unsupervised image super-resolution methods. Context-wise, Network, Fusion, Data Augmentation, Multi-task Learning, Network Interpolation, and Self-ensemble are the recent advancements integrated with image super-resolution.
Popular domain-specific applications of deep learning-enabled image super-resolution are depth Map Super-resolution, Face Image Super-resolution, Hyperspectral Image Super-resolution, Real-world Image Super-resolution, and Video Super-resolution. Literature surveys on deep learning-based image super-resolution present image super-resolution approach, important issues, publicly available benchmark datasets, performance evaluation metrics, and future directions.