基于几何约束的文本图像倾斜角检测算法

针对文本图像倾斜检测的问题, 提出了一种新的基于几何约束的文本图像倾斜角自动检测算法。该算法采用边界标记自动机的方法对一组同行字符轮廓进行检测从而得到该组字符轮廓的最低点信息, 再用矩的方法剔除噪声字符, 并确定页面的倾斜角度。实验结果表明, 该算法在检测效率与准确率上都有了明显的提高, 同时在处理较大倾斜角和较少字符数目的倾斜检测中也有较好的执行效率。因此, 该算法可广泛应用于包括英文、中文、日文在内的多种语言文本图像的倾斜检测中。

Document image skew detection algorithm based on geometric constraint

For the problem of detecting the angle of the document image, this paper proposed a new skew detection algorithm based on the geometric constraint of the document image. The algorithm got the lowest pixels of a set of characters by the method of region-labeling-automata. It evaluated the skew angle by applying moment calculation, and described the method to get characters in one text-line in detail. The experimental results show that the efficiency and the accuracy are both improved by this algorithm, what's more, the document images with large skew angle and the document images containing a few characters can also be treated by this algorithm. Therefore, the document images of various languages, including English, Chinese and Japanese can be validly treated by this algorithm.