基于模糊技术的文本检测提取方法研究

Method for text detection and extraction in images based on fuzzy techniques

基于内容的索引和检索是实现巨量数字视频资料有效检索与浏览的一种重要方法,是当今的研究热点之一,而检测提取其中的文本是实现该目标的一个重要途径。文章提出一种检测提取图像中文本的新方法,其特点在于应用模糊技术对子图像变异灰度直方图(VGH)中的文本映射特征进行识别,从而实现检测定位图像中的文本;在讨论子图像VGH中文本映射特征性质和阈值参数选取算法的基础上,着重研究了模糊技术用于文本映射特征识别问题,给出了具有鲁棒性的文本检测提取算法;实验结果表明,新方法切实可行,相关算法正确有效。

Content-based indexing and retrieval is an important method for effectively retrieving and browsing the huge amount of digital video information,and it is one of popular research issues nowadays. In this paper, a new method for text detection and extraction is presented, the characteristic of which is the application of fuzzy techniques to recognizing the text mapping features in subimage variant gray-scale histogram（VGHs） and implementing the text area detection and location. The properties of text mapping features in subimage VGHs are analyzed first. Then the key techniques and their related algorithms are discussed in detail, including how to choose the parameter λeffectively, how to detect features in VGHs of a subimage quickly and how to apply the fuzzy techniques to recognizing effectively the text mapping features in VGHs. In the end, some experimental results are given showing the validity of the algorithms discussed above.