Abstract

The Hough transform (HT) is an efficient tool for parametric curve detection and therefore plays an important role in digital image processing. Realtime implementation of the HT using digital electronic computers is restricted to very small images as a result of memory and computation time requirements. These restrictions can be overcome by performing the HT optically and taking advantage of the inherent parallelism of optics. The HT maps each point or small line segment in the image plane into a curve in the parameter domain. A curve in the image plane can be subdivided into many small line segments, each of which is mapped in the parameter domain into a curve of different parameters depending on the location of the small line segment. The coordinates of the intersection of these curves can be determined by thresholding. We implemented the HT optically using a space-variant holographic filter, which is a 2-D array of optically recorded Fourier transform holograms of the parametric curves introduced from a liquid crystal spatial light modulator under microcomputer control. Arrays of 60 X 60 holograms were realized; however, arrays of 256 X 256 holograms are feasible. Experimental results for the detection of straight lines in normal and Cartesian parametrization and of circles are presented. Potential applications for 2-D and 3-D image processing and pattern recognition are discussed.

© 1986 Optical Society of America