A Shape-Adaptive Thinning Method for Binary Images

摘要

Binary images thinning has wide applications in image processing, machine vision, and pattern recognition systems. A new method of shape-adaptive thinning algorithm and its implementation is presented in the paper. It is mainly used for obtaining skeletons of binary images especially, handwritten and printed characters for OCR or other image processing and recognition systems. The proposed thinning algorithm is based on the substitution of some pixels on the strokes or curves. To achieve this three groups of templates are designed according to the complexity of these stroke and curve connectedness. The experimental results show the algorithm preserves well the original structure information of input binary images, which is specially helpful for the recognition of characters such as Chinese characters, English alphabets and numerals.