Blind Face Restoration Survey
Abstract
This survey provides a comprehensive review of deep learning methods for blind face restoration—recovering high-quality face images from degraded inputs without knowing the degradation type. We analyze GAN-based, diffusion-based, and hybrid approaches.
Topics Covered
- GAN-based methods: StyleGAN priors, GFPGAN, CodeFormer
- Diffusion models: Recent advances in restoration
- Benchmark datasets: CelebA-HQ, FFHQ, LFW
- Evaluation metrics: PSNR, SSIM, FID, LPIPS
Related Topics
📄 Access:
Google Scholar