The course will be split into two five-week sections.
The first half will focus on 3D understanding and recognition from inverse light transport.
1. Intro - cameras, lenses, sensors
2. Image Formation - radiometry, light transport, dynamic range
3. BRDF - reflectance maps, specularities
4. Material Recognition - microfacets, subsurface scattering
5. Global versus local shading models - inter-reflections, low-dimensional light transport
6. Algorithms: color constancy (retinex), photometric stereo, shape from shading, shape from texture
The second half will focus on multiview geometry.
1. Rigid body motion - rodrigues formula, quaternions
2. Camera models - radial distortion, perspective projection
3. Projective Geometry - conics, planar homographies, overview of stratification
4. Camera Calibration - autocalibration
5. Mathematics of multi-view geometry - essential matrix, fundamental matrix, epipolar geometry, 8-point algorithm
6. Estimation - least squares, bundle adjustment
7. Practical Concerns - feature point detection, description, RANSAC
8. Projection of smooth surfaces - differential geometry, occluding contours, aspect graphs
9. Structure from motion - rigid/multi-body/non-rigid factorization
10. Stereo reconstruction - dynamic programming, graph cuts, structured-light, multi-view