手写汉字变形问题是非特定人脱机手写汉字识别中的关键同题。本文从预处理的角度探讨了直接从汉字点阵图像上矫正手写变形、减小属于同一类别的不同模式之I町差异的途径，提出了一种新的非线性规一化方法。该方法同时考虑了汉字笔画像素点和背量空白像素点，汉字点阵中的所有像素点均具有笔画密度描述，且将不同的像素点赋予不同的笔画密度，笔画密度时描述具有二维性、局部性，使得笔画密度的描述更为合理，规一化后的汉字点阵中笔画的分布更加趋向均匀．实验结果证实了本文方法的有效性。-The problem of handwritten Chinese character deformation is the key problem of non- specific person　s offline handwritten Chinese character recognition. In this paper, we discuss the way of correcting the handwriting deformation directly　the Chinese character lattice image and reducing the difference between the different patterns belonging to the same category　the pretreatment point of view, and propose a new nonlinear normalization method. This method also takes into account the pixel of Chinese characters and the white pixels of the back volume. All the pixels in the Chinese word lattice have the descr iption of the stroke density, and the different pixels are given different stroke density. The stroke density is described as having two dimensions , Localized, making the descr iption of the density of strokes more reasonable, standardized distribution of Chinese characters in the distribution of strokes more evenly evenly. The experimental results confirm the effectiveness of the p