One of the most challenging tasks in optical examination of natural fibers – besides measuring fiber lengths and diameters – is the determination of the possible existence of branches in a fiber. Since usual automatic fiber examination systems are known to be irritated by fiber bifurcations, possibly leading to unreliable length measurements, a mechanism to detect and evaluate side branches would be helpful for both the developers and users of automatic fiber examination systems. Additionally, fibers with higher and lower amounts of bifurcations can be expected to show different mechanical properties when spun to a yarn or used in a fiber -reinforced plastic composite. Thus, different statistical methods have been used for image processing of scanned or microscopic pictures of natural fibers, such as hemp, flax, or linen. One of them, the so-called random-walk method, has been used before for the examination of microscopic images of yarns [1] and knitted fabrics [2-4], resulting in a novel method to measure yarn hairiness and fiber fineness optically. The article gives an overview of different shapes of single hemp fibers and fiber bundles and shows which differentiations can be made using statistical methods on the scanned pictures, pre-treated with newly developed numerical filters. These approaches are used to distinguish between hemp fibers which have been pre-treated with different methods.