Title :
The Hough transform for locating cell nuclei
Author :
Thomas, A.D.H. ; Davies, T. ; Luxmoore, A.R.
Author_Institution :
Univ. Coll. of Swansea, UK
Abstract :
The motivation in selecting the Hough transform as a means for locating nuclei was its ability to reveal any `round´ objects in an image. A fast method for detecting ellipses, and hence circles, has been described. This method was applied to a number of typical cervical smear images. The inherent robustness of the Hough transform when dealing with obscured or deformed shapes resulted in few nuclei being missed, and some were detected which were not instantly obvious to the naked eye. The transform failed only in cases of severe deformation, or when nuclei had an eccentricity well outside the sought-after range. Few artifacts were mistakenly identified
Keywords :
computerised pattern recognition; computerised picture processing; medical diagnostic computing; transforms; Hough transform; cervical smear images; evidence-gathering procedure; image segmentation; locating cell nuclei;
Conference_Titel :
Applications of Image Processing in Mass Health Screening, IEE Colloquium on
Conference_Location :
London