Text Line Based Correction of Distorted Document Images

Distortion usually appears in document images captured by digital camera or scanner. In this paper, a text line based method for efficient calibration of document images is presented. Image bridging is used to generate points at lower edges of text lines. With these points, the least-square fitting method is used to get curves or line sets of each text line. Based on the curves or line sets, offsets of lower edge points of each text line can be calculated. According to linear interpolation, offset of each pixel in document images is obtained. And then the document image is rectified using offsets of pixels. This method does not require extra devices, regardless of the con- tent of document. The effectiveness and robustness of the method proposed in this paper is proved by experimental results.