Convolutional Neural Network (CNN) is a specialized class of deep neural networks designed for image and grid-based data processing, utilizing convolutional and pooling layers to automatically learn hierarchical features for tasks like image classification and object detection.

The Generative Adversarial Network (GAN) demonstrated remarkable success in image inpainting by seamlessly reconstructing missing parts of an image, as well as in super resolution, enhancing the finer details of low-resolution images with astonishing clarity.

We utilize CNN-based phase estimation to accurately restore images with distorted phases in medical imaging, and subsequently employ GANs to restore values of unsampled pixels in sparsely sampled images, facilitating high-definition upscaling.

Image restoration, Inpainting, and Upscaling by CNN and GAN

Image restoration by Convolutional Neural Network (CNN)

Image Inpainting and Upscaling by Generative Adversarial Network (GAN)