摘要
视频图像中的文本包含了丰富的语义层次上的内容描述信息,为基于语义的图像检索提供重要的索引信息资源.提出了一种基于Homogeneity和支持向量机(support vector machine)的视频图像中文本检测方法,首先将图像由空间域映射到Homogeneity域中,然后对映射到Homogeneity空间中的图像进行特征提取,利用SVM判别文本区域.实验表明此文本检测方法优于用基于边缘特征的文本检测方法.
Text data presented in images and video contains useful and important semantic information for automatic indexing. In this paper, a method for text detection based on homogeneity and SVM is proposed. First an original image is mapped from space domain to homogeneity domain, and then text region property is confirmed by SVM trained to extract property feature in homogeneity domain. Comparison with the text detection method based on edge features shows that the proposed method has a better accuracy.
出处
《智能系统学报》
2007年第1期69-73,共5页
CAAI Transactions on Intelligent Systems
基金
国家自然科学基金资助项目(60573071).
