支票手写体汉字大写金额识别的非线性规范化

针对银行支票图像大写金额的无限制手写体汉字识别问题，进行了基于密度均衡原则的非线性规范化研究。提出了一种改进的非线性规范化方法．该方法定义的基于笔画间距和宽度的密度函数，不仅能较好地克服笔画变形的局部性、不规则性，而且能使同一字符内以及不同字符之间的笔画粗细趋于一致；同时，确定了图像中字符的有效区域，并据此改进了基于密度均衡原则的通用表达式，有效地解决了字符整体倾斜和单个笔画比较突出的问题，实验结果表明：该方法比其他同类方法效果更佳，可使银行支票图像的大写金额识别系统的识别正确率提高约1．5％。

An Improved Nonlinear Normalization Method and Its Application to Handwritten Legal Amount Recognition on Chinese Cheque

An improved nonlinear normalization method based on density equalization of the exact character area is proposed for recognition of unconstrained handwritten Chinese characters on an image of Chinese cheque. By the method, the density of strokes, calculated based on the thickness of the stroke and the space between strokes, can minimize the localization and irregularity of the shape distortions of the characters. Strokes in different characters are also normalized to have the equal thickness. By modifying the expression of density equalization according to the exact character area, the problems of global incline and highlighting of certain strokes are effectually rectified. Experiments show that the nonlinear normalization method can get better shape correction effect than other methods. Application of this method to handwritten legal amount recognition to Chinese cheques showed the rate of correct recognition is improved by around 1.5%.