基于压缩域的JPEG图像检索技术  

An Image Retrieval Technique in JPEG Compressed Domain

作　　者：黄帆[1] 

机构地区：[1]中山大学计算机科学系,广州510275

出　　处：《现代计算机》2006年第5期75-76,99,共3页Modern Computer

摘　　要：目前大部分图像都以压缩的形式来存储和传输,其中以JPEG格式更为流行。检索时对JPEG图像进行完全解压会相当费时。为此,本文提出了一种基于压缩域的JPEG图像快速检索的技术方案,从图像的主要DCT(离散余弦变换)系数中提取关键信息进行计算得到图像的特征向量,并与目标图像的特征向量比较后得出差异度较小的图像作为检索结果。这种检索方法缩短了检索的时间,图像的匹配精度较高。Most images are now stored and transferred in the compressed form, of which JPEG format turns even more popular. However, it takes much time to decode the JPEG images fully while indexing them. This paper proposes a fast image retrieval technique in JPEG compressed domain. It extracts key information from the important DCT（Discrete Cosine Transform） coefficients of the image and gains the feature vectors through calculating.These feature vectors can be compared with those from the target image so that the differences between them can be found. In this way, images that have relatively small differences with the target are to be the output. This technique makes the indexing time shorter and the precision of matching higher.

关 键 词：图像检索 压缩域 特征向量 JPEG 

分 类 号：TP391.41[自动化与计算机技术—计算机应用技术]