中文摘要：

为解决复杂拍摄背景下采集到的文本图像难以快速、准确地进行倾斜校正的问题,提出一种新的倾斜文本图像校正方法。该方法通过对选定图像的投影窗口的水平投影统计分析计算,自适应地筛选出只含有文字的文本特征子区。在特征子区内部,通过采用连通域横向搜索法定位文字行并拟合直线计算出文本线的倾斜角度,从而得到文本图像的倾斜角度,最后利用改进的快速旋转变换算法进行倾斜校正。实验结果表明,该倾斜校正方法不受成像背景、复杂版面的影响,具有速度快、精度高、适应性强的优点,可广泛应用于对速度和效率有更高要求的移动终端平台上。

英文摘要：

In order to solve the problem that the text images captured in complex background are difficult to correct tilts quickly and accurately,we propose a new correction method for tilted text images.Based on statistically analysing and calculating the horizontal projection of projection window in selected image,the method adaptively screens out the feature sub-region of texts only containing characters.Inside the feature sub-region,the method positions the characters line by using the connected domain lateral search approach and fits the straight line for calculating the tilt angle of text line,therefore obtains tilt angle of text image.Finally,it employs the improved fast rotation transform algorithm for tilt correction.Experimental results show that the tilt correction method is not affected by imaging background and complex layout,and has the advantages of high speed,high precision and strong adaptability,it can be widely used in mobile terminal platforms which have higher requirements on speed and efficiency.