Title detection algorithm for OCR credentials image based on directional projection and divergence
Zhiwen Wang, Guoqing Xie, Shaozi Li, Zhi Liu · 2010
According to the OCR character region in OCR credentials image, a new skew detection method was presented by using Bresenham's integer only line drawing algorithm and within-group measure of dispersion. Several directional projection are calculated in the algorithm by using Bresenham's integer line drawing algorithm. At the same time calculating minimal within-group measure of dispersion of the directional projection. After searching minimal value of all within-class divergence, the angle relative of minimal value is the one of the OCR credentials. Experiment results show that the new method can measure out the slanting degree of an OCR credentials image efficiently and accurately.