Degradation regression with uncertainty for blind super-resolution

Shang Li, Guixuan Zhang, Zhengxiong Luo, Jie Liu, Zhi Zeng, Shuwu Zhang

Abstract: Highlights•A probabilistic degradation estimator (PDE) is proposed to estimate the degradation as a certain distribution rather than a fixed point. It is more robust to the estimation error and could help synthesize degradation-specific training samples.•An IoU-based degradation regression loss with uncertainty is proposed. It can optimize PDE more effectively by making the predicted degradation parameters and associated tolerances more correlated.•Extensive experiments on both synthetic and real-world images show that our degradation estimator can help the SR model achieve better performance in various cases.