论文部分内容阅读
笔迹鉴别多用匹配方法比较字符的书写风格,而字符图像的预处理和归一化对匹配是非常重要的。本文介绍笔迹鉴别的字符图像预处理和一种形状匹配方法。预处理主要介绍二值图像的噪声消除和归一化方法。噪声消除的方法是平滑、轮廓跟踪和填充。为保持字符中的书写特征,点阵的归一化是线性的,但字符位置和尺度的确定非常重要。本文给出了三种归一化方法:四边定界法、重心对准法和单边定界法,并在此基础上用图像匹配方法进行书写人识别的实验。匹配方法是通过距离变换快速实现的。实验结果表明,重心对准归一化最适合于笔迹鉴别问题,距离变换匹配得到的识别率也比较令人满意。
Handwriting identification methods are more used to compare the writing style of characters, while character image preprocessing and normalization is very important for matching. This article introduces the handwriting recognition character image preprocessing and a shape matching method. Pretreatment mainly introduces the noise reduction and normalization method of binary image. The methods for noise cancellation are smoothing, contour tracking and padding. The normalization of the lattice is linear in order to preserve the writing features in the characters, but the determination of the character’s position and scale is very important. In this paper, three normalization methods are given: the four-edge delimitation method, the centroid alignment method and the unilateral delimitation method, and on the basis of this, an image matching method is used to test the writer’s recognition. Matching method is quickly achieved by distance transformation. The experimental results show that the normalization of the center of gravity is most suitable for handwriting recognition, and the recognition rate of distance transform matching is also satisfactory.