Volumetric Super Resolution

Volumetric super-resolution aims to enhance the resolution of three-dimensional images, particularly in medical imaging and scientific visualization, where anisotropic resolutions (different resolutions in different directions) are common. Current research heavily utilizes deep learning, employing architectures such as convolutional neural networks (CNNs), generative adversarial networks (GANs), transformers, diffusion models, and implicit neural representations (INRs) to achieve this upscaling. These methods are being evaluated and improved upon using both synthetic and real-world datasets, focusing on mitigating artifacts, improving inter-slice consistency, and handling arbitrary upsampling factors. The improved resolution offers significant benefits for medical diagnosis, scientific analysis of complex data, and other applications requiring high-fidelity 3D visualizations.