摘要： 为解决复杂背景图像中的文字检测问题，提出一种基于数学形态学的复杂背景文字检测方法。采用多尺度多结构的数学形态学边缘检测算子抽取图像边缘，通过由粗到精的策略进行文字检测。实验结果表明，该方法能够提高文字检测的准确率和召回率，同时缩短文字检测时间。

关键词: 文字检测, 边缘检测, 数学形态学, 多尺度, 支持向量机, 连通元标定

Abstract: To detect text from complex background image, a text detection approach based on multi-structure element of morphology is proposed. It adopts a multi-structure element of morphology-based scheme to extract contrast edges. Then a coarse to fine text detection strategy is performed to detect text lines. Experimental results demonstrate that the proposed method can not only improve precision rate and recall rate but also save detection time.

Key words: text detection, edge detection, mathematic morphology, multi-scale, Support Vector Machine(SVM), connected component labeling

中图分类号: 

• TN919.81