数字图书馆图像资源检索框架的构建与实现——基于非下采样的Contourlet变换

The Construction and Implementation of Image Retrieval Framework for Digital Library——Based on the Non Sampling Contourlet Transform

图像检索为数字图书馆的发展提供了技术支持,图书馆应重视数字化发展以提升服务质量。本文提出一种数字图书馆图像资源检索框架,并对系统的实现过程做了详细的分析。同时,在提取图像特征时提出了一种基于非下采样的Contourlet变换图像检索算法(NSCT),能够在大量图像数据中挖掘有效的特征信息。该算法首先对图像进行多尺度、多方向分解,然后计算低频和高频中不同方向的子带系数的标准差和均值作为图像的纹理特征。实验结果显示,本文提出的图像检索框架具有可行性,能够为用户提供更优质的搜索服务,并且与同类特征提取算法进行比较,该算法具有良好的检索性能和较高的查准率、查全率。

Image retrieval has provided a technical support for digital library. [ Method/process] This paper proposed a digital library image resource retrieval framework and analyzed the implementation process of the system in detail. At the same time, an image retrieval algorithm based on non sampling contourlet transform was proposed, which could discover valuable infor- mation from large image database. Firstly, it applied NSCT to decompose images into multi - scale and multi - directional ones, and then calculated mean value and standard deviation of low - frequency and high - frequency subband coefficient n^atrices as im- age texture features. [Result/conclusion] Experimental results showed that the image retrieval framework proposed in this paper was feasible and could provide better search service for users. The methods had a higher retrieval performance, precision and ac- curacy than the similar algorithms.