一种基于版面结构距离的文档图像检索算法

Document image retrieval method based on layout distance measures

介绍了一种基于版面结构距离的文档图像检索算法,使用版面特征作为文档图像的特征检索图像。先将文档图像进行梯度和最大梯度差(MGD)计算,然后使用MGD值作为一个窗口对文本区域进行融合,将文档图像以行线的形式标示出来。同时给出了检索的匹配方法,并对匹配方法进行了实验。实验结果表明,该检索方法具有较高的查准率,具有很好的抗倾斜和抗缩放效果。

This paper presented a method for document image retrieval by layout analysis as a search feature.At first,compute the gradient and the maximum gradient different （MGD）for the input document image,then the MGD values are filled within a local window to merge text segment.The line of document image is marked in the form of line.This paper gives the mach algorithm between the two images,tests the algorithm by image skewing and scaling.The experiment indicates that the algorithm is good at precision and robustness.