新的文本图像倾斜检测及校正算法

New document skew detection and correction algorithm

在文档扫描过程中,文档可能会发生倾斜,而很多字符识别和布局分析算法都对倾斜十分敏感,文本图像的倾斜检测及校正就成为文档分析不可缺少的环节。提出了一种新的倾斜文本图像的校正方法,该方法首先获取文档图像的boun- ding box,以bounding box面积最小作为倾斜校正的最终目标,并使用遗传算法搜索该最小值。实验结果表明,该算法对倾斜角的检测具有较高的精确度。

During document scanning, the input document images inevitably introduce skew, while the algorithms for layout analysis and character recognition are sensitive to page skew. A document skew detection and correction based on GA algorithms is proposed. Firstly, the bounding box of the skew document is gotten, by using the characters of the bounding box, then GA is used to detect the skew angle. The experimental results show that the proposed algorithm guarantee the accuracy for skew document.