一种改进的基于颜色-空间特征的图像检索方法  

An Improved Image Retrieval Method Based on Color-Spatial Feature

作　　者：方俊[1] 郭雷[1] 汪子强[1] 

机构地区：[1]西北工业大学自动化学院,西安710072

出　　处：《计算机工程与应用》2005年第25期68-70,87,共4页Computer Engineering and Applications

基　　金：国家自然科学基金(编号:60175001)

摘　　要：颜色量化是基于颜色图像检索的一个热点。由于量化边界处的颜色具有连续性和相似性,文章提出了一种改进的基于模糊量化的颜色量化方法,以减小量化误差,使量化方法更接近于人的主观视觉感知。基于这种量化方法,提出了一种基于颜色—空间特征的检索算法,采用了一种相似度量方法以利用相同直方图区间内的像素统计与空间信息之间的相关性。实验结果表明该方法具有较高的检索有效性。Color quantization is a focus of image retrieval based on color feature.To continuity and similarity of the color of quantization boundary,an improved color quantization method based on obscure quantization is proposed to minimize color quantization error,which is close to human＇s subjective vision perception.Based on the color quantization method,this paper proposes an image retrieval algorithm based on color-spatial feature,a similarity measure method is adopted to utilize the correlation between statistic and spatial information of pixels in the same histogram bin. Experimental results show the method has significant high retrieval effectiveness.

关 键 词：颜色量化 模糊量化 图像检索 颜色-空间特征 

分 类 号：TP391[自动化与计算机技术—计算机应用技术]