摘要
提出了一种基于DCT压缩域的字符定位算法,能够快速定位出具有复杂背景图像中的字符区域。该算法面向部分解码后的JPEG图像,从Y分量DCT压缩码流中提取出一种新的字符/非字符分类特征,并采用自适应阈值法实现分类,利用投影法确定出字符区的位置。实验表明,该算法对不同复杂背景下的JPEG图像,可以有效实现中、英文字符区的提取,查全率和查准率可以达到90%以上,处理速度快,能够实现实时处理。
A fast and efficient automatic text location method is presented. Text regions are segmented from JPEG compressed images with complex background us- ing a new feature, which is extracted from the DCT- compressed domain. Hence only a very small amount of decompressing operations is required. Then a projecting and merging algorithm is used to locate the final text ar- eas. Experimental results show that this method works well on various language text locations with precision and recall of more than 90%.
出处
《测控技术》
CSCD
2005年第5期48-51,共4页
Measurement & Control Technology
基金
国家自然科学基金资助项目(60402036)
北京市基金资助项目(4042008)
教育部博士点基金资助项目(20040005015)
关键词
DCT系数
字符定位
压缩域处理
加权频率
自适应阈值
DCT coefficient
text location
com- pressed-domain processing
weighted frequency
adaptive threshold
