Photometric stereo from a single fixed viewpoint captures the intrinsic reflectance properties of surfaces under varying illumination. This approach estimates surface normals and albedo while relying on controlled lighting changes rather than camera motion.
By leveraging a calibrated light source configuration, the method turns each fixed light direction into a linear constraint that can be solved in closed form. The result is a practical pipeline for shape and material reconstruction without complex hardware.
| Key Parameter | Typical Value or Setting | Effect on Results | Practical Guidance |
|---|---|---|---|
| Number of Light Directions | 3–9+ directions | More directions reduce noise and ambiguity in normals | Use non-coplanar lights for full rank estimation |
| Surface Albedo Assumption | Lambertian or near-Lambertian | Violations introduce errors in shape and reflectance | Validate with known materials or extend model |
| Camera Calibration | Intrinsic known, extrinsic fixed | Incorrect calibration distorts normal estimation | Use a calibration target and bundle adjustment if possible |
| Illumination Consistency | Stable color temperature and intensity | Shifts in lighting can corrupt reflectance estimates | Control environment or estimate light directions online |
Lighting Configuration and Calibration Strategy
Accurate photometric stereo depends on precise knowledge of light directions in the camera coordinate system. You can measure light directions physically using a calibrated sphere or estimate them jointly with surface properties.
Place lights at known positions around the object, and avoid symmetry configurations that cause degenerate solutions. A carefully designed lighting setup improves robustness and reduces shadow and interreflection artifacts.
Surface Normal Estimation from Linear Inverse
Under Lambertian assumptions, the relationship between observed intensities, light directions, and surface normals is linear. Solving this linear system yields per-pixel normals and albedo directly from multiple images.
Handle non-Lambertian regions by masking or regularizing, and ensure numerical stability by checking condition numbers of the lighting matrix. Smoothing normals after estimation preserves coherence while removing spurious noise.
Shape Reconstruction and Coordinate Alignment
Integrating estimated normals across the surface produces a watertight representation of shape when the field is well-behaved. However, gradients measured in image space must be aligned consistently with the camera reference frame.
Use integration techniques such as Poisson reconstruction or multi-scale methods to recover height maps. Align the reconstructed mesh to the camera by registering control points or known references for metric accuracy.
Material Classification and Reflectance Modeling
Beyond albedo, photometric stereo can feed features into classifiers that distinguish between matte, glossy, or metallic surfaces. Extract local cues such as sharpness of highlights and response across light directions.
Combine these cues with domain knowledge to label materials, and integrate with a material database for better scene understanding. Keep lighting conditions consistent when comparing materials across objects.
Key Takeaways and Recommended Workflow
- Use a minimum of three well-spaced, non-coplanar lights to avoid degenerate solutions.
- Calibrate the camera and verify light directions with a reference sphere before full reconstruction.
- Prefer Lambertian surfaces or validate deviations with extended photometric models.
- Check conditioning of the lighting matrix to ensure stable normal estimation.
- Post-process normals with filtering and integrate gradients carefully for reliable shape.
- Evaluate material responses across light directions to support classification tasks.
- Document lighting setup and calibration parameters to enable reproducibility across scenes.
FAQ
Reader questions
How many light directions are needed for reliable photometric stereo from a single fixed viewpoint?
Three non-coplanar directions provide a minimum solution, but five to nine directions significantly reduce noise and ambiguity in both shape and reflectance estimates.
Can photometric stereo handle interreflections and cast shadows in a single fixed view setup?
Standard linear methods assume direct illumination only; interreflections and shadows typically introduce bias, but robust variants and preprocessing masks can mitigate these effects in controlled environments.
What role does camera calibration play when lighting is precisely controlled?
Accurate intrinsic calibration and stable extrinsic positioning ensure that estimated normals correspond to the true surface geometry, preventing systematic distortions in reconstructed shape. Yes, you can optimize light directions jointly with surface normals by treating them as unknowns under Lambertian assumptions, provided observations from multiple lighting settings are available and the lighting configuration is sufficiently diverse.