基于核心区域检测的手写维吾尔文本行倾斜校正方法

Detecting Core Region Based Handwritten Uygur Text Line Skew Correction Algorithm

为了提高手写维吾尔文本图像的可读性与识别精度,对其倾斜校正显得非常重要。本文以克服传统Hough方法需要手动设置不同参数、计算量繁重等缺点为出发点,在原始启发式学习方法的基础上,确定文本行核心区域、下基线,最终完成倾斜校正,采用大津算法计算阈值,确定文本行核心区域,而后使用最小二乘拟合技术得到文本行下基线;每个倾斜角在合理间隔范围内,对图片采用旋转算法完成倾斜校正。这种技术首次运用在手写维吾尔文本行的倾斜校正中,初步得到了非常明显的实验结果。

In order to improve the readability and the automatic recognition of Uygur handwritten document images,preprocessing seems to be very important. In this paper,as a starting point to overcome the shortcomings in traditional method of Hough,it need to set different parameters manually and the calculation is heavy. On the basis of the original heuristic learning methods,to determine the the core region and lower baseline for text line,and finish the skew correction. In this paper,we use the algorithm of Otsu to calculate the threshold,determine the text line core region,and obtain the lower baseline by the technique of least-squares fit. For each angle in a reasonable interval,a rotating algorithm is applied to finish the skew correction of the image. This technique that we use in handwritten Uygur documents for skew correction is the first time,and the experiment results is very obvious.