自相似特征彩色图像检索算法  

Color Image Retrieval Algorithm Using Self-similar Characteristic

作　　者：许相莉[1] 周春光[1] 于哲舟[1] 张利彪[1] 孙昊翔[1] 

机构地区：[1]吉林大学计算机科学与技术学院,吉林长春130012

出　　处：《小型微型计算机系统》2008年第5期945-949,共5页Journal of Chinese Computer Systems

基　　金：国家自然科学基金项目(60433020;60673099)资助;教育部"符号计算与知识工程"重点实验室项目(02090)资助;国家"九八五"工程"计算与软件科学科技创新平台"项目资助

摘　　要：针对传统的分形图像检索方法不能很好的体现图像的颜色自相似特征,且具有计算量大、速度慢的缺点,提出了适合彩色图像的自相似特征编码方法,将编码空间由灰度空间扩展到了彩色空间,充分体现了图像的颜色自相似特征.为了达到更好的检索效果,提出了对图像的自相似特征编码进行奇异值分解来构造特征向量的方法,并定义了相似度度量公式来支持图像检索.实验结果表明,本文提出的算法具有较好的可行性和有效性.Traditional fractal methods used in image retrieval have disadvantages that can＇t incarnate the colory self-similar characteristic well, and have large calculation and low velocity, in order to resolve these problems, a self-similar characteristic coding method is proposed, which expands the code space from gray space to the color space, and can incarnate the colory self- similar characteristic effectively. In order to achieve a better retrieval effect, feature vector is constructed through singular value decomposing the self-similar characteristic code, and a similarity formula is defined for supporting the image retrieval. Experiments showed that the proposed algorithms are feasible and valid.

关 键 词：自相似特征 奇异值分解 图像检索 

分 类 号：TP391.41[自动化与计算机技术—计算机应用技术]