Segmentation-Driven Monocular Shape from Polarization based on Physical Model
Researchers have developed a new method for monocular shape-from-polarization (SfP) that addresses the issue of azimuth angle ambiguity. This technique segments images into smaller regions, allowing for more accurate local shape recovery and preserving surface continuity. By incorporating a multi-scale fusion convexity prior, the method enhances the reconstruction of fine details and demonstrates improved accuracy on synthetic and real-world datasets compared to existing physics-based SfP approaches. AI