基于纹理的图像字符自动定位技术对比研究

Comparative study on texture-based text location techniques in images

随着信息技术的迅速发展,基于内容的图像检索技术引起了研究者的广泛关注。自动抽取图像/视频的语义内容是图像/视频检索中研究的重点和难点。本文通过分析多种字符定位技术,给出了基于纹理分析的字符定位算法的一般框架。对各种基于纹理分析的字符定位方法进行了对比研究,尤其深入地研究了纹理特征提取,分类器设计及边界位置确定对定位结果的影响。最后通过预分类的方法对算法进行改进,实验结果表明,该方法在保持正确定位率基本不变的情况下,可以大幅度提高处理速度。

With the development of information technique, Content-based accessing to images has captured the attention of many researchers in recent years. Extracting the semantics present in images and videos is the difficulty and the key point of images retrieval. A general flow graphic for textual-based text location algorithm is proposed. Comparative study on textual-based location techniques, especially on the effects on result that produced by feature extracting, classifier and location method is given. Finally, an efficient text location method is addressed with the improved speed of algorithm.