Computer Vision

Computer vision studies how to extract geometric and semantic structure from images. The field’s two grand questions — where is this in 3D? and what is this? — drive the two halves of this section: classical geometry-based methods (sections 1–6) and modern representation-learning methods (sections 7–8).

The geometric pipeline flows roughly:

1. Image Formation — how scenes become pixels.
2. Image Processing — per-pixel operations: filtering, gradients, edges.
3. Image Transformations — 2D geometric warps: affine, projective, homographies.
4. Feature-Based Alignment — detecting, describing, and matching keypoints across images.
5. 3D Image Geometry — the pinhole model, stereo, and epipolar geometry.
6. Multi-View Reconstruction — Structure from Motion and Multi-View Stereo.

Modern methods continue:

7. Neural Fields — NeRF and related volumetric representations.
8. Image Generation — diffusion and generative models.

Cross-Domain Connections

• 4. Visual Servoing — the robotics bridge that closes the perception-action loop using the pinhole model and the robot Jacobian.