Detection of linear, low-contrast tracks in noisy images via optical Hough Transform
Shih‐Chun Lin, R. Scott Boughton · Annual Meeting Optical Society of America · 1989
The Hough transform is a pattern recognition technique for identifying straight lines in a 2-D picture. With this technique, any straight line or collinear line segments in the object domain are transformed to a point in the feature domain, thus allowing a higher degree of detectability. For uses requiring real time processing such as industrial product inspection and track detection, several optical implementations have been proposed. In this paper we report detection of low-contrast linear tracks in noisy images by a simple optical Hough transform setup. Films containing noisy 512×512 digital images, each with a 2-pixel wide track, were used as inputs. The output spot was detected by a linear CCD array. Both coherent light and incoherent light were tried. It was found that with coherent light the outputs were completely swarmed by speckle noise. With incoherent light, the signal-to-noise ratio of the output depends on the contrast of the track to the background.