Graph-based checkerboard indexing and its applications in camera calibration and multi-view processing

In this thesis, a new method for checkerboard indexing is presented. A graph-based approach is used to capture the topology of the checkerboard pattern. The regions and crosspoints of the pattern are identified using image segmentation and represented by a graph. At this point, no explicit mathematical model is assumed which describes the distortion in the checkerboard image. A special calibration pattern design allows the identification of a unique reference point in the pattern. This way, the crosspoints can be interpreted as the grid points of a target coordinate system. Possible applications in the fields of camera calibration and multi-view processing are discussed.