Dear Imrus Salehin,

For image reconstruction, particularly in medical imaging, GANs can be effective, with several variants such as CycleGAN, Pix2Pix, SRGAN, and WGAN-GP showing good performance for reconstruction tasks. However, newer models like Diffusion Models, Transformers for Image Generation with Diffusion, and Vision Transformers (ViT) are often better choices for achieving high-quality and stable reconstructions.

Here are some references you may find helpful:

For GANs:

For Diffusion Models:

For Vision Transformers (ViT):

1. https://github.com/lihkinVerma/Image-Reconstruction-by-Vision-Transformer

Kind Regards,

Md Foysal Ahmed