Under the microscope, scratches or abrasions on hard otherwise flat surfaces are usually revealed as straight-lined irregularities. At a more macroscopic level creases in thin sheets such as of paper and textile fabrics are also observed to be straight-lined. A computer-aided image analytical method is described here not only for identifying such features but also for counting them, measuring their lengths and evaluating their contrast. Further measures are derived that are in accord with the qualitative visual impact of each line within the milleau of lines in the original image. The method makes use of a parametric transformation from two orthogonally-illuminated images of the surface using the equation p=x∙cos(θ) + y∙sin(θ) where x,y are image coordinates, θ is the angle that a straight line makes with the x-axis and p is the perpendicular distance of that line from the coordinate origin. As distinct from the well-known Hough transform, estimates are made for θ at all points in the initial images that are illuminated at a low angle from two orthogonal directions.
Swift, J. Alan
"The Detection and Quantification of Straight-Lined Irregularities on Surfaces,"
Scanning Microscopy: Vol. 1992
, Article 26.
Available at: https://digitalcommons.usu.edu/microscopy/vol1992/iss6/26