摘要
网络动画中往往包含丰富的字符信息,如果能够将这些字符信息加以提取和识别,将对网络动画的有效过滤具有重要意义。论文介绍了一种新的基于类边缘文本提取算法及其实现,该算法利用DCT方法提取出字符的类边缘信息,然后采用基于映射方法进行定位。实验结果表明,该方法能够准确、有效地定位和提取网络动画中的文字区域。
Animations in the webpage usually contain text region, and if such text information can be extracted and recognized, the accuracy and performance of animation, understanding, retrieval and classification will be greatly improved. This paper presents a novel text detection algorithm based on edge-like feature, including its implementation. The proposed algorithm adopts DCT transform to obtain the edge-like features and then the projection-based method is used for text localization. Experimental results indicate that the proposed algorithm can localize and extract the text region from animations effectively and accurately.
出处
《信息安全与通信保密》
2007年第11期66-67,70,共3页
Information Security and Communications Privacy
基金
国家自然科学基金项目(60402019)
教育部新世纪优秀人才支持计划项目(编号NCET-06-0393)
关键词
边缘检测
DCT变换
文字定位
动画理解
edge-like detection
DCT
text region localization
animation understanding
