Catalogue of Artificial Intelligence Techniques
Structure from Motion
Keywords: rigidity assumption
Author(s): R.M. Cameron-Jones
The calculation of three-dimensional global structure from the two-dimensional properties of features visible in a time series of images. The features are usually points, but may be lines, etc. Most methods require the correspondence of features between images and rely upon the rigidity assumption, i.e., that the environment is rigid. There are many results on the number of views required for a minimal solution to exist for a given number of points visible in all the images, amongst which the best known (first derived for orthographic projection by Ullman under the rigidity assumption) is that three views of four points are required. In a rigid environment, the method is a form of stereo--the determination of structure from images taken from more than one viewpoint.
- Faugeras, O.D., Lustman, F. and Toscani, G., Motion and structure from motion from point and line matches Proceedings of the First International Conference on Computer Vision