一种基于笔划宽度的印鉴分割方法

印鉴分割是正确识别真假印鉴的关键问题，本文给出了一种新的印鉴图像分割算法。该算法基于样本章（印鉴）和待验章（印鉴）在具有相同的笔划宽度时，分割效果最佳，最利于真假印鉴的识别。选定一定范围的灰度值作为可能的阈值，从中选 出使待验章和样本章的笔划宽度相差最小的阈值作为最佳阈值。对位于待验章和样本章的交集中的边缘点分别在待验章和样本章中求水平和垂直方向上的笔划宽度，并且仅取宽度差最小的一个方向上的宽度差，很好地避免了印泥的浓淡、盖印时用力的深浅、印泥污染、笔划断裂和分裂带来的干扰。实验和实践证明，该方法分割效果好，并具有很好的稳定性，在识别真假印鉴的实践中得到了很好的应用。

A Method for Seal Image Segmentation Based on Stroke Width

Seal image segmentation is a key problem in anti-forged seal recognition. This paper proposes a new method. Based on the fact anti-forged seal recognition, the method can get the best performance when the testing seal image and model seal image have the same stroke width. It selects a threshold from a certain range to regard as one threshold which is the best threshold just as it makes the difference of stroke width in testing and model seal image minimum. Using edge points in intersection set of testing and model seal imaged to calculate stroke width in horizontal and vertical directions, and choosing the minimum difference of stroke width in one direction as the difference of stroke width for one point, the new method can avoid disturbances caused by deep or shallow ink, strong or weak press seal, broken or polarized strok, stroke pollution of ink. Experiments prove that the method can get good performance and is very stable.