SFMambaNet: Spectral-Frequency Enhanced Selective State Space Model for Correspondence Pruning
Researchers have introduced SFMambaNet, a novel network designed for correspondence pruning in computer vision. This model integrates spectral-frequency domain perception with Mamba-based architecture to better identify inlier correspondences. SFMambaNet utilizes a Local Spectral-Geometric Attention block and a Spectral-Integrated Global Mamba block to enhance feature discriminability and suppress noise accumulation, outperforming existing state-of-the-art methods. AI
IMPACT Introduces a novel architecture for correspondence pruning, potentially improving accuracy in computer vision tasks.