SFMambaNet: spectral-frequency SSM for correspondence pruning

Arxiv CS.CV (computer vision)

SFMambaNet enhances selective state space models with spectral-frequency features to improve inlier identification in correspondence pruning. The method outperforms GNN-based approaches in distinguishing subtle geometric differences.

mambastate-space-modelcomputer-vision

Arxiv CS.CV (computer vision)·Zhihua WangYanping LiYizhang Liu

7 days ago