Photometric Stereo for General BRDFs via Reflection Sparsity Modeling.

Area of Science:

• Computer Vision
• Computer Graphics
• Computational Imaging

Background:

• Photometric stereo reconstructs surface shape from images under varying illumination.
• Modeling complex Bidirectional Reflectance Distribution Functions (BRDFs) remains a challenge.
• Existing methods often struggle with general BRDFs, limiting their applicability.

Purpose of the Study:

• To propose a pixelwise photometric stereo method for object surfaces with general BRDFs.
• To develop a robust reflection modeling approach capturing diffuse, specular, and shadow characteristics.
• To enhance the accuracy and applicability of photometric stereo for diverse materials.

Main Methods:

• A graph-based approach models light directions, capturing reflection characteristics.
• Diffuse reflection is modeled via local variations, specular reflection via group sparsity.
• Shadows are modeled using weighted l1-norm, with the optimization problem formulated as second-order cone programming.

Main Results:

• The method effectively handles object surfaces with general BRDFs, including isotropic and anisotropic materials.
• Experimental results on synthetic and real-world scenes demonstrate superior performance compared to state-of-the-art methods.
• Accurate surface reconstruction is achieved by appropriately modeling reflection components.

Conclusions:

• The proposed photometric stereo method offers a robust solution for 3D surface reconstruction with complex BRDFs.
• The reflection modeling strategy effectively integrates diffuse, specular, and shadow properties.
• This work advances the capabilities of photometric stereo for a wider range of materials and applications.