OPTICAL FEATURE EXTRACTION VIA THE RADON TRANSFORM.

The authors show that by first measuring the line integrals of a two-dimensional picture f(x,y) via the Radon transform, certain feature-extraction operations useful in pattern recognition may be rapidly computed. In particular, they consider these features: (1) moments of f(x,y) invariant to translation, rotation, geometric scaling and linear contrast scaling; (2) two geometric features, polar projections and convex hull, and (3) Features of the Fourier power spectrum of f(x,y) (4) the Hough transform for detection of edges. Much of the motivation for this work lies in its implementation as an electro-optical pattern recognition system. Some experimental results are shown.