面向OSD选单文字分割算法的研究

Study of text segmentation algorithm for OSD menu

针对现有的自动化OSD选单图像文字分割存在亮度不均匀、图形噪声等问题,提出了一种面向OSD选单的文字分割算法,首先结合顶帽变换和改进的双阈值Otsu法对图像进行光照校正和图像二值化;通过垂直投影分析和Canny算子边缘检测法去除图形噪声,结合二分法和投影法解决文字黏合问题,最终达到OSD选单文字分割的目的。实验结果表明,提出的算法能够解决OSD选单亮度不均匀和图形噪声的问题,文字分割正确率达到88%以上,比传统投影法的正确率提高了67%。

For automation OSD menu text image segmentation exiting uneven brightness, pattern noise and other issues, a text seg- mentation algorithm for the OSD menu is put forward, combined with Top-Hat algorithm and improved dual-threshold OTSU meth- od for image illumination correction and image binarization; then uses the improved projection method to remove pattern noise, and uses dichotomy to solve the problem of text cohesion, and finally completes text segmentation. Experimental results show that this method can solve the problems of uneven brightness and pattern noise. The accuracy rate of text segmentation reaches over 88%, increasing by 67% compared with the traditional projection method.