Shading-Based Surface Detail Recovery Under General Unknown Illumination

Area of Science:

• Computer Vision
• 3D Reconstruction
• Image Processing

Background:

• Reconstructing 3D object shapes from multi-view images under unknown illumination is a fundamental computer vision challenge.
• High-quality reconstruction is difficult, especially with fine details and non-uniform object albedo.
• Existing methods often overlook the importance of surface orientation and illumination modeling.

Purpose of the Study:

• To introduce a novel approach for recovering surface details and 3D object shapes from multi-view images.
• To effectively model and account for unknown, general illumination conditions.
• To improve the quality and robustness of 3D reconstructions, particularly for objects with fine details and varying albedo.

Main Methods:

• Introduced vertex overall illumination vectors to model illumination effects.
• Developed a total variation (TV) based approach for recovering surface details using shading and multi-view stereo (MVS).
• Formulated reconstruction as a constrained TV-minimization problem, solving unknowns simultaneously using an augmented Lagrangian method.

Main Results:

• The proposed method demonstrates robustness and stability in 3D shape reconstruction.
• Achieved efficient recovery of high-quality surface details, even from coarse initial models.
• Successfully reconstructed challenging objects with varying albedo under complex illumination.

Conclusions:

• The TV-regularized approach effectively recovers surface details and 3D shapes by simultaneously addressing illumination and geometry.
• This method offers significant improvements over traditional multi-view stereo techniques, especially for complex scenes.
• The approach is well-suited for detailed 3D reconstruction tasks where illumination is unknown and object surfaces vary.