Depth super-resolution from explicit and implicit high-frequency features

Xin Qiao, Chenyang Ge, Youmin Zhang, Yanhui Zhou, Fabio Tosi, Matteo Poggi, Stefano Mattoccia

Abstract: Highlights•DSR-EI employs an efficient transformer for explicit, HF feature extraction.•We propose LCF that can obtain accurate, implicit HF information.•We propose AFFM to counter the information loss issue.•DSR-EI outperforms other SoTA methods.