Evaluation of Skeletonization Methods for Document Images with Rotation States

Images skeletonization is an important process for many applications of pattern recognition. Currently, lot of skeletonization methods have been proposed to dealwith usual skeletonization challenges such as nosy contour, topology preserving, and two-pixel thinness problems. Unfortunately, no one among them focused in producinga fix skeleton with different rotation states. This paper presents an evaluation of set of recent and well-knownskeletonization methods for binary document images to deal with the rotation challenge. We implemented and tested the Bataineh, Abu-Ain, Huang, and K3M methods overdocument images with several rotations. The DIBCO2010, H_DIBCO2010_GT benchmark dataset with benchmark measurements are used to evaluate the performance of the involved methods. The experiments showed a various results based on the adopted measurements.