Fiber identification in microscopy by ridge detection and grouping

Abstract

In microscopy, a common task is the identification of individual objects having some particular shape, after which various features can be measured and feature statistics taken over the set of objects. The identification process can be automated by applying appropriate computer vision techniques. The author addresses the specific problem of fiber identification. Fibers appear as thin lines or curves in an image. In a 3D graph or 'surface plot' of the image, they would appear as ridges or valleys. The paper describes a method of finding fibers based on the detection of individual ridge 'edgels'; grouping of these edgels into simple, generally non-overlapping, curves; and the further grouping of curves into extended fibers.<>