The thinning methodology is novel in terms of its ability to incorporate character shape specific knowledge while constructing the thinned skeleton. But removal of spurious strokes or shape deformation in thinning is a difficult problem. In this paper, we have proposed a novel medial-axis based thinning strategy used for performing skeletonization of noisy character images. The proposed algorithm produces segmented strokes in vector form as a by-product. Hence further stroke segmentation is not required. Experiment is done on printed English, Bengali, Hindi, and Tamil characters and we obtain less spurious branches compared to other thinning methods without any post processing. We have concluded with a proposed methodology to extract structural features from thinned character images. This feature set improves the performance of existing OCR for Indian languages. © 2010 IEEE.