Key research themes
1. How can shape-from-shading methods be enhanced to resolve ambiguities and improve 3D surface recovery under varying lighting and material conditions?
This research area addresses the inherent ill-posedness and ambiguities in traditional shape-from-shading (SFS) formulations, particularly focusing on mathematical well-posedness, incorporation of advanced reflectance models, and user-guided optimization strategies to improve robustness and accuracy of 3D shape recovery from shading cues.
2. How can image-based cues beyond shading, such as occlusions, silhouettes, and structural priors, be integrated to reconstruct high-quality 3D shapes from single images?
This line of investigation combines classical shape-from-shading with other monocular cues and novel priors to overcome the limitations of shading alone. It also incorporates data-driven and structural constraints particularly suited for complex objects such as human faces and hair, aiming for dense, detailed, and realistic 3D reconstructions from minimal input.
3. What advances in differential geometric and mathematical modeling provide new frameworks for representing and reconstructing shape and texture from images?
This research theme focuses on the use of differential geometry, curvature analysis, and advanced mathematical representations to extract shape and texture information from images and video. It also encompasses theoretical and practical methods that support robust 3D reconstruction through mathematical regularization, curvature-based segmentation, and texture synthesis.