Gradient Space and Reflectance Map | Photometric Stereo

Gradient Space and Reflectance Map | Photometric Stereo

by Shree Nayar

Formal & Physical Sciences AI — Models & Research UYQArtificial intelligence (AI)UYQVComputer vision
🎙 Shree Nayar 👥 96K 📅 March 21, 2021 ⏱ 14 min 👁 15K 📄 tutorial 🧭 2026-08-17
Available in: English (current) Français

Keywords

gradient spacereflectance mapphotometric stereosurface normalLambertian

Summary

This lecture introduces the concept of gradient space as a two-dimensional representation of surface orientation, defined by the partial derivatives of depth. The surface normal is expressed as (p, q, 1) and its unit vector is derived. The gradient space is visualized as a plane at z=1, where each point corresponds to a unique orientation. The reflectance map R(p,q) is then defined as a function that maps surface orientation to image intensity, given a known light source direction and surface reflectance model. For a Lambertian surface, the reflectance map simplifies to the cosine of the incidence angle, expressed as the dot product of unit normal and source direction. Iso-brightness contours in the reflectance map are shown to be conic sections, resulting from the intersection of a cone of normals with the gradient plane. The lecture concludes by highlighting the ambiguity in recovering surface shape from a single image, as a single intensity value corresponds to an entire contour of possible orientations, motivating the need for multiple light sources in photometric stereo.

171 words

Critical Evaluation

Value of the Information & Strength of the Argument

The lecture provides a clear and rigorous explanation of fundamental concepts in photometric stereo. It builds the argument step-by-step, starting from the definition of gradient space, then deriving the reflectance map for Lambertian surfaces, and finally illustrating the ambiguity in shape recovery. The use of geometric visualizations (cone-plane intersections) effectively conveys the relationship between surface orientation and brightness. The mathematical derivations are accurate and well-motivated, making the content valuable for students and practitioners. The argumentation is solid, with no logical gaps or unsupported claims.

93 words

Title / Content Match

The title accurately reflects the content, which focuses on gradient space and reflectance maps as foundational concepts for photometric stereo.

Quality & Reliability

9/10

Lecture by a leading expert in computer vision, based on established mathematical models (Lambertian reflectance, gradient space). The presentation is clear, rigorous, and well-structured, with derivations and visualizations. No unsupported claims; the content aligns with standard computer vision literature.

Key Moments

Markers derived by PSI from the transcript: the creator did not define chapters.

Contribution & Novelties

The lecture provides a clear and accessible introduction to gradient space and reflectance maps, which are foundational for photometric stereo. It effectively bridges geometric intuition and mathematical formulation, making these concepts accessible to learners. The visualization of iso-brightness contours as conic sections is particularly illuminating.

Pour aller plus loin :

91 words

Radar Profile

The radar profile shows high scores across all dimensions, indicating a well-balanced and reliable educational resource. The strong scores in quality and reliability reflect the expert presentation and accurate content, while the high technical level indicates depth suitable for an intermediate audience.

Reliability 9/10