摘要： 在现有的汉字方向分解网格特征基础上，对其分析和改进，提出了一套新的汉字特征方案：首先对原汉字进行轮廓提取，在轮廓汉字上进行有重叠的矩形弹性网格划分，然后对原汉字用边缘检测算子做边缘梯度方向角分解，得到横、竖、撇、捺4 个方向的子图像，再把网格应用到这4 个方向的子图像上，统计每个网格中的黑像素占这个子图像总像素的比例，4 个子图像的所有网格的比例值构成的矢量就是汉字的特征。使用不依赖于分类器的特征标准比较了新旧方案，表明新方案优于旧方案。

关键词: 汉字方向分解网格特征；矩形弹性网格；边缘梯度方向角

Abstract: A new feature of Chinese character based on analyzing and improving the directional decomposition cellular feature of Chinese character is proposed, which uses elastic rectangular mesh with overlap to divide boundary of character into meshes and uses edge gradient angle to decompose the original character into directional sub-patterns. Testing based on a classifier-irrelevant feature measure shows this new feature is better than the old directional decomposition cellular feature

Key words: Directional decomposition cellular feature of Chinese character; Elastic rectangular mesh; Edge gradient angle