blul sky transfer to dusk sky.
AI auto generate sky for sky transfer.

an unsupervised methdod using GAN can simultaneously preserve the rich spectral information of multi- spectral images and the spatial information of panchromatic images.

This study investigates a reconstruction method designed to generate 7T magnetic resonance images (MRI) from 3T MRI based on a generative adversarial network. Compared with current reconstruction methods, this method can simultaneously reconstruct well-defined anatomical details and salient blood vessels. This reconstruction method may be useful for increasing the efficiency of brain MRI examinations.

Results of 7T MRI reconstruction using FocalGAN on two testing pairs shown in coronal view. 3T MRI (A) is the input, 7T MRI (B) is the reference, Out (C) designates the reconstruction generated by FocalGAN, error maps (D) are shown in the rightmost column. Regions outlined in red are expanded beneath the relevant panel (E-G). Maximal intensity projection (MIP) images show improved imaging of blood vessel reconstruction in two test pairs. 7T MRI MIP (A) refers to reference images, Out MIP (B) designates results generated using FocalGAN. Transferring FocalGAN to domain transfer. Before segmentation, We transfer 3T MRI to 1.5T MRI, which can greatly improve the robust of our segmentation model (10% improvements).

MRI denoise in a supervised way using dncnn.
CT denoise in an unsupervised way using cyclegan.

FusionGAN enables final fused image simultaneously keeps the thermal radiation in a infrared image and the textures in a visible image and avoids manually designing complicated activity level measurements and fusion rules as in traditional methods.