摘要： 由照相机拍摄的文档图像可能因扭曲变形导致OCR软件不能正确识别。为解决上述问题，采用图像分割技术进行单词及文本线检测，利用线性拟合得到单词的较低基线和较高基线，根据校正基线对单词进行旋转和垂直位移，得到校正后的图像。实验结果表明，该方法能快速有效地校正扭曲的文档图像，使校正后的图像在光学字符识别阶段的识别率有较大提高。

关键词: 文档图像扭曲校正, 图像内容分割, 校正基线

Abstract: Non-linear warping often appears in document images which is captured by the camera. In order to solve the problem, this paper uses image segmentation technology to detect words and text lines, applies linear fit to get lower baseline and upper baseline of the words, and makes the words rotation and vertical displace according to upper baseline and lower baseline, so that the corrected image can be obtained. Experimental results indicate that the method can rectify distorted images quickly and effectively, and elevate the probability of the rectified image identification in the optics character identification stage.

Key words: document image distortion correction, image content segmentation, correction baseline

中图分类号: 

• TN911.73