Chapter 10 - Variational Multiview Reconstruction Shape optimization Shape optimization: find 3d shape that best corresponds to the given images. Explicit shape representations For example, splin...

Chapter 9 - Introduction to Variational Methods Motivation: Optimization-Based Methods let you specify what you want from a solution directly. This can be much better than “cooking up” an algorithm...

Chapter 8 - Direct Approaches to Visual SLAM SLAM = Simultaneous Localization And Mapping Indirect vs. Direct methods Classical Multi-View Reconstruction Pipeline The classical approach tackles st...

Chapter 7 - Bundle Adjustment & Non-Linear Optimization Reconstruction under Noise Linear approaches are usually prone to noise. We now assume that $\tilde{x}_1, \tilde{x}_2$ are noisy data po...

Chapter 6 - Reconstruction from Multiple Views Not just two, but multiple views. Each new view gives 6 new parameters, but many more point measurements -> ratio params / measurements improves. ...

Chapter 5 - Reconstruction from Two Views Assumptions two consecutive frames given: set of corresponding points static scene (no movements during camera motion) intrinsic camera parameter...

Chapter 4 - Estimating Point Correspondence Goal: find point descriptors/characteristic image features to be able to identify keypoints accross several images from different views. Things to consi...

Chapter 3 - Perspective Projection Goal of MVG: invert the image formation process. One part of the formation process is the camera motion (last lecture). The second one is the projection of point...

Chapter 2 - Representing a Moving Scene Origins of 3D Reconstruction 3D reconstruction is a classical ill-posed problem, as its solutions are not unique (most extreme example: imagine a photograph...

These notes are lecture notes on the Computer Vision II - Multiple View Geometry course held in the summer term 2021 by Prof. Daniel Cremers/Prof. Florian Bernard. Ref Chapter 1 - Mathematical Ba...