3D Computer Vision: Lecture 1 Part 2 - Projective Geometry

Added:

Conformal Geometric Algebra (CGA)

While classical projective geometry using homogeneous coordinates is the standard framework for 3D computer vision, Conformal Geometric Algebra (CGA) offers a powerful alternative mathematical language. Projective geometry struggles to naturally represent and transform certain geometric primitives—such as circles, spheres, and angles—and often requires complex matrix representations. In contrast, CGA represents points, lines, planes, circles, and spheres as algebraic elements within a unified vector space. It simplifies geometric operations by replacing coordinate-heavy matrix transformations with coordinate-free 'rotors' (generalizations of quaternions). Proponents of CGA argue that it provides a more intuitive, covariant, and computationally elegant framework that avoids the limitations of projective duality and coordinate-dependent formulations, offering a cleaner foundation for 3D vision, kinematics, and robotics.

Linear Algebra fundamentals, particularly matrix-vector multiplication, determinants, and systems of linear equations.
Basic Cartesian coordinate geometry, including standard representations of 2D points and lines.
The concept of homogeneous coordinates and how they represent points and lines at infinity.
Basic understanding of 2D matrix transformations, such as scaling, rotation, and translation.
3D Projective Geometry, extending 2D concepts to 3D space, including planes, quadrics, and 3D homographies.
Camera models and calibration, mapping 3D world points to 2D image coordinates using projection matrices.
Epipolar Geometry and Multiple View Geometry, exploring the geometric relations between multiple camera views (e.g., Fundamental and Essential matrices).
Practical image stitching and warping, applying homography estimation and RANSAC algorithms to create panoramas.
Structure from Motion (SfM) and Simultaneous Localization and Mapping (SLAM) for real-time 3D reconstruction.
15.3K views124likes1:13:14@cvrplabatnus8618Original Release: 2021-01-20

In projective 2D space, conics (parabolas, ellipses, hyperbolas) are represented by homogeneous equations x^T C x = 0 with five degrees of freedom, requiring five points to uniquely define them; degenerate conics occur when the matrix rank drops below 3, representing either a single line (rank 1) or two intersecting lines (rank 2). Projective transformations (homographies) are invertible 3×3 matrices preserving collinearity, forming a hierarchy from Euclidean transformations (preserving distance and angles) through similarity transformations (preserving shape) to affine transformations (preserving parallelism), with the most general projective transformation having eight degrees of freedom and being computable from four point correspondences.