Return
CoDiFSR: Code Diffusion paradigm with prior knowledge distillation for face super-resolution
DOI:10.1016/j.patcog.2026.113745.png)
Abstract
En 中文
• A novel Code Diffusion paradigm with two-stage training protocol to exploit prior guidance. • Extract highly compact Facial Prior Embeddings as prior guidance for FSR backbone. • Well-constructed Facial Prior Codebook with expressive and high-quality FPEs. • Leveraging knowledge distillation for Code Diffusion model to refine the low-quality FPE.
Keywords:
Code Diffusion
Facial Prior Embeddings
Face Super-Resolution
Knowledge Distillation
Prior Guidance
Journal
IF:
7.6
Papers:
1.3W
Citations:
4.5W

