DRNet: Learning a dynamic recursion network for chaotic rain streak removal

Zhiying Jiang, Risheng Liu, Shuzhou Yang, Zengxi Zhang, Xin Fan

Abstract: Highlights•The interlaced rain is independently processed with a recursive mechanism.•We propose the dynamic cross-level recruitment for feature compensation.•Contrastive learning is introduced to advance the generalization in real-world.•Extensive comparison and practical evaluation demonstrate the superiority of DRNet.