针对现有基于图像文档转换为文本后进行文档检索的方法,无法满足当今超大量数字图像库的处理场景。文中提出一种基于文本布局块的文档图像检索方法。根据文本布局块之间的距离特征,定义了新的距离函数,利用新的距离函数计算得到文本布局块之间的距离矩阵,并结合匈牙利算法求出文档图像的最佳匹配结果。通过大量实验证明,所提方法能够有效地提高图像文档检索准确度,并且能保证78.2%的正确率。
The existing methods of document retrieval based on the conversion ot linage documents into text can not meet the processing scenes of today's large number of digital image databases. This paper proposes a document image retrieval method based on text layout block. According to the feature of distance between text blocks, a new distance function is defined. Then, the new distance function is used to calculate the distance matrix between text blocks. Finally, the best matching result is obtained by combining the Hungarian algorithm. A lot of experiments show that this method can effectively improve the image document retrieval accuracy, and can guarantee the correct rate of 78.2%.