3D Computer Vision

Flashcards covering key concepts in 3D Computer Vision from the lecture notes.

What is the main focus of multi-view reconstruction in computer vision?

Reconstructing 3D scenery from multiple views or images taken from different angles.

What is the Tomasi-Kanade factorization?

A method used to factorize a measurement matrix in multi-view reconstruction into motion and structure components.

What are the key steps in the bundle adjustment process?

Optimization of camera parameters and spatial coordinates to minimize projection errors.

What is the role of the trifocal tensor in three-view geometry?

It is an extension of epipolar geometry used to relate 3D points and lines across three images.

What is meant by 'weak-perspective projection' in the context of camera models?

An approximation where the depth of scene points is assumed constant across images, simplifying the projection calculations.

How does one estimate similarity transformations between point clouds in stereo reconstructions?

By analyzing common points across the clouds and solving for scale, rotation, and translation parameters.

What numerical method is commonly used in bundle adjustment?

The Levenberg-Marquardt algorithm.

What issue does Tomasi-Kanade factorization with missing data address?

It modifies the factorization process to deal with scenarios where some feature points are not visible in all images.

What is the significance of singular value decomposition (SVD) in factorizations?

It's used to reduce noise and retain important features by keeping only the largest singular values.

Why is the center of gravity important in the orthogonal projection?

It simplifies the coordinate system and the equations used for reconstruction.