摘要： 针对传统视频文字检测方法存在效率较低、计算复杂、精确度不高的不足，提出一种视频文字检测方法，通过基于边缘密度和连通域分析的文字粗检测得到候选文本行，利用稀疏表达分类产生的过完备词典进行文字行验证。实验结果表明，该方法具有较好的检测性能，可以应用于视频检索系统。

关键词: 文字检测, 边缘密度, 连通域分析, 稀疏表达分类

Abstract: Aiming at the shortages of low efficiency, computational complexity and accuracy not high in traditional video text detection method, this paper presents a video text detection method based on sparse representation classification. It utilizes edge intensity and connected-components analysis algorithm to get raw text lines, and verifies them with non-complete dictionary generated by sparse presentation classification. Experimental results show that this method has better detection performance, and it can be applied to video classification and retrieval system.

Key words: text detection, edge intensity, connected domain analysis, sparse representation classification

中图分类号: 

• TP391.9