DCD用于相关反馈的自然图像检索

Relevance Feedback-based Natural Image Retrieval Using DCD

基于内容的图像检索是多媒体应用研究领域的一项关键技术。传统的CBIR系统一般采用基于图像的低级物理特征进行特征匹配,MPEG-7标准的提出为图像内容特征的提取与描述提供了一种新型简便的方法,本文讨论了应用MPEG-7的主体颜色描述符DCD对图像特征提取和表示的方法,并在图像检索中的应用研究基础之上,提出了在使用相关反馈的图像检索中的改进方案,实现了基于DCD特征描述的自然图像检索。对于1350幅的数字图象组成的数据集进行了实验,并对实验结果进行分析,大部分结果是令人满意的,提高了基于内容的图像检索的精度。

Content-based image retrieval is a key technique in multimedia. Normally, traditional CBIB system describes and compares images by using low-level physical features, and the international standard MPEG-7 is the novel method whichcan extract and express content features effectivdy. In this paper, MPEG-7 Dominant Color Descrilxor （DCD） is used to extract and express image features. It puts forward a new retrieval method using DCD, and proposes techniques to improve the effectiveness of Relevance Feedback （BF） based image retrieval using DCD. The retrieving is achieved. The data set composed of 1350 digital images is experimented with, and the experiment results have been analyzed, some of which are satisfied, and the precision based on content image retrieving has been improved.