摘要
多媒体数据库应用需要有效的基于内容相似性检索方法。颜色特征由于其简单、计算复杂度低及对几何变换的不变性成为机器可自动提取的图像内容中最重要的特征。文章讨论了颜色特征的表示及其进展。直方图是使用最普遍的颜色特征描述符,它必须选择与人类视觉机制一致的颜色空间和量化模式。直方图与空间关系的组合可提高图像内容描述的精度,因而提供更好的颜色特征匹配。由于基于小波变换的编码技术已成为JPEG2000等图像编码标准的核心,因此基于小波变换系数特征描述方法已越来越受到重视。由于图像颜色的心理作用可影响观察者对图像的理解,如何建立与心理活动及视觉机制相适应的颜色特征模型,提取这些语义级的高级抽象内容是我们必须面对的挑战。
The application of Multimedia database requires efficient content-based similarity retrieval method.Due to its simple,low computation complexity and invariant to geometric transform,color feature is the most important feature in automatic extractable image contents.Color feature description and its advance are discussed in this paper.Histogram is the most popular color feature description.combination of histogram and space relation such as local histogram,CCV(Color Consistent Vector) can increase image description precision,so as to provide better solution to color feature match.Due to Wavelet transform based coding technology have become core of image coding standard such as JPEG 2000,so wavelet transform coefficient based feature description have get more and more attention.A true content based image retrieval system require semantic based features.Image color can effect viewer's emotion,how to extract these high abstract content is a challenge that must confront.
出处
《计算机工程与应用》
CSCD
北大核心
2005年第16期48-50,113,共4页
Computer Engineering and Applications
