Currently, biological image analysis generally relies on manual segmentation of objects and subsequent quantification of desired features. In the case of filamentous structures, such as cytoskeletal structures, the natural clutter and image variability impose significant challenges to automated segmentation methods. In this work, we describe a method to address these issues. Starting from a point on the target structure, we examine potential directions to extend the trace along the structure. The final trace of the object is selected from the candidates as the best match with the image data. Our method can be used in collecting statistical features from images of filamentous structures. We demonstrate the method on microtubule image sequences, determining tip positions in image sequences, calculating curvature values along the microtubule body, and estimating microtubule tracks over time. Given the increased analytical efficiency of our method, we can now acquire sufficient data to employ statistical analysis methods on biological image sets. Additionally, we can now examine quantitative data beyond the traditional manual measurements. For example, bending tendency can be quantified as a function of curviness observed in images, which is impractical to capture manually. Statistics derived from this parameter, such as the spatial distribution of highly curved objects, may raise novel biological questions and hypotheses. In the case of microtubules, curvature may be an invaluable measure in quantifying cell shape changes or changes in the direction of cellular or growth cone migration. Our algorithm can be easily modified to measure various spatial and temporal characteristics to all sorts of filamentous structures.