A new technique for determining the distortion parameters (location, orientation, and scale) of general 2-D objects is introduced. It uses the straight-line Hough transform as a feature space. The technique is very efficient and robust, since the dimensionality of the feature space is low and since it uses input images directly (with no preprocessing such as segmentation). Because the feature space allows separation of translation and rotation effects, a hierarchical algorithm to discriminate among objects and to detect object rotation and translation using projections and slices of the Hough space is possible.
© 1988 Optical Society of America
Raghuram Krishnapuram and David Casasent, "Hough transform projections and slices for object discrimination and distortion estimation," Appl. Opt. 27, 3451-3460 (1988)