Directional image analysis wi the Nough and Radon transforms
Rancarn M. Rancayyan, William Alexander Rolston · 1998
The ohjecrivz of the method to be dcvclopcd m thia paper is to obtair quantitatwe information on the dircclional distribulion of features in givcn image The Hough transform em be regarded as 0 method of transformmg nn mngo into domain that reduces the image mto set of parameters trom which it is ensw to ohtam the desired information ~n the nnage. The main drawback of the I-luugh tmnaform is tint it is usually pzllurmed on binary Images, and la thus dcpsndenl on Lhc binarisetion method that ra used in 5egrnznting the Image. The implementation proposed In this paper avoids thtr dl-awback by deitgnmg modfied vetwon of the Hough transform that wes concepts from the Radon transtom to obtain HuughIRadon transform. In the plopored algorithm, the ~magc is tint transfolmed inlo the HoughRadon domain, wlrk at the sanrc lme second rh~dow domain is created to keep count or thc number of pmls at part~culal- parameter point. The HoughiRadon-tmnshrmed image rs then filtered to obtam coherent peaks Thz shadow parameter space la then antegrated to ubraln count of thr number of p~xels curresponding to dtrectiond features ~n each angle band. Due to the summine of intcnslties in the Radon mnsCorrn step, the d~orithm dctacts direLtiunill clcmznta on the baa& of oricriw~ tion as well a inteni;ity.