Machine extraction of textured polygons from arbitrary backgrounds

This thesis has dealt with the problem of extracting polygonal shapes from an arbitrary background.

Shape, size and texture content are essential discriminators for extraction. An edge image is obtained by applying the Sobel operator on the digitized picture. This edge image serves as the primary sketch and is further processed by the normalized Hough transform to obtain shape information. The final step involves decisions based on shape, size and texture content.

The basic approach is extensively evaluated through experiments involving license plate recognition. It has been demonstrated that it is feasible to extract a license plate from a view of the rear of a vehicle, automatically.