Random Sampling Methods for Two-View Geometry Estimation

Random Sampling Methods for Two-View Geometry Estimation

TNO Defensie en Veiligheid
den Hollander, R.J.M.

2007

The automatic interpretation of a scene in an image is the ultimate goal in computer vision. It is an ambitious goal, since the aim is essentially to create a human-like system. Humans, as we all know, have a remarkable ability in processing and interpreting visual information. We can make the transition from the pictorial information to the meaning of the image content in milliseconds. The scene is decomposed into separate objects and each is compared to collected information from past experiences. Without a serious effort we are able to identify the objects and reason about their structures. The structure of objects is essential in the recognition process. Our experience tells us in what way structures may change when viewed under different conditions. For example, an object may be imaged from another viewpoint or the amount of illumination may change. Furthermore, other objects may partially occlude the object of interest. All these factors should be taken into account when an automated system, i.e. a camera connected to a computer, is designed to interpret or recognize a scene. etc

Vision

http://resolver.tudelft.nl/uuid:db857bb6-bfff-4313-b7bb-a7c70aeb0263

224091

Technische Universiteit, Delft

9789090218397

ASCI dissertation series

Proefschrift Technische Universiteit Delft, 2007

doctoral thesis