This paper proposes an automatic measurement algorithm for calculating statistics of pulp fiber lengths and curvature. The fibers are extracted from digital images, and a parametric model is fit to the extracted objects. The modeling approach improves the robustness of the detection and enables connecting broken fiber parts into full length curves. The developed algorithm is compared to an alternative method, and concluded to yield similar results while providing additional features, such as curvature estimates, of the detected fibers.