区域相关融合纹理特征FDPC图书馆文档图像检索

为实现数字图书馆馆藏资源的检索效率和检测精度的提升,提出基于快速纹理密度极值的聚类算法(FDPC)的图像检索策略.首先,对图书馆图像资源检索问题进行基于内容的检索框架构建,然后采用直方均衡以及中值滤波策略实现图像资源的背景处理和噪声过滤,并通过二值化对图书馆馆藏图像资源进行处理,获得检索框架图像输入的预处理操作;其次,利用极值密度聚类算法(DPC)对图像的分类问题进行研究,同时为了增强算法的聚类效果,基于动态距离截断策略对其进行改进,获得DPC算法性能的有效提高;最后,基于图书馆标准测试库(Corel)对所提算法的性能进行实验验证,试验结果表明所提方法具有更高的检索效率和检索精度.

Regional Correlation Fusion Texture Feature Based on FDPC Algorithm for Library Document Image Retrieval

In order to improve the accuracy of image retrieval in digital library,a new method of regional correlation fusion texture feature based on FDPC algorithm for library document image retrieval algorithm has been proposed.Firstly,the content of library document image retrieval framework has been presented,and the document images conducted for deposing and background processing based on median filtering and histogram equalization strategy,and then the image binary processing achieved to realized input image pre processing.Secondly,the density peak clustering (DPC) has been used to classify the images,and to improve the clustering effect,and the convergence performance of DPC algorithm been improved by using dynamic range truncation method.Finally,the validity of the proposed retrieval method is verified through the experimental comparison of the library Corel test database.