基于可变模板的文本倾斜检测

在对文本扫描输入的过程中，文本图像不可避免地会发生倾斜，倾斜校正将为图文分割、文字识别等后续处理工作创造良好的条件。基于可变模板技术，提出一种新的倾斜检测方法。在构造表征扫描文本的可变模板和定义合适的能量函数的基础上，采用遗传算法进行快速优化搜索。实验结果表明，该方法能够精确检测文本的倾斜角度，并且不受倾角大小和文字方向的影响，具有较强的抗噪声性和较快的收敛速度。

Document Skew Detection Based on Deformable Template

In process of document scanning input, images may be skew somehow. Severe image skew makes next step image segmentation difficult and lowers character recognition accuracy. A new approach of skew detection based on deformable template is presented in this paper. Genetic algorithm is used to search optimal solution on the basis of constructed document template and defined energy function. Experiment results show the method can detect the skew angle efficiently and accurately in a few minutes and bear high noise endurance and application adaptability.