Methods for Structure from Motion

Structure from motion, the problem of estimating 3D structure from 2D images hereof, is one of the most popular and well studied problems within computer vision. In part because it is academically interesting, but also because it holds a wealth of commercially very interesting prospects, e.g. within entertainment, reverse engineering and architecture.This thesis is a study within this area of structure from motion. The result of the work, which this thesis represents is the development of new methods for addressing some of the problems within the field. Mainly in robustifying the factorization approach, relaxing the rigidity constrains, and in considering alternative ways of solving the surface estimation problem.In Danish:Structure from motion problematikken beskæftiger sig med at estimere 3D struktur fra 2D afbildninger heraf. Denne problemstilling er en af de mest populære og velstuderede inden for coputer vision. Dette skyldes tildels, at den er akademisk interessant, men også at den har et stort kommercielt potientiale.Denne afhandling er rapporteringen af et studie inden for dette område, structure from motion. Dette studie har resulteret i udviklingen af nye metoder til at imødekomme nogle af de problemer, der er inden for området. Hovedsagligt drejer dette sig om at gøre de såkaldte faktoriserings metoder mere robuste, at undersøge hvorledes stivheds antagelsen kan blødes op samt at undersøge alternative måder at løse overflade-estimerings-problemet på.