摘要
在基于颜色的图像相似比较中,传统的处理方法大多采用各子空间的概率作为特征,在此基础上定义图像的距离和相似度.但这些方法可能导致色彩比对失真的情况出现;在同一类的石材中颜色和纹理有很大的差异,甚至会出现两张分属异类的石材图像在单纯从颜色、纹理等特征上进行判断比同类很多石材图像更为"相似".文中引入了先对石材的样本图像进行分类,定义图像与各大类的相似度.在类内定义较为一般性的相似度.综合二者给出图像的整体相似度定义,实验表明该方法具有较好的分类效果和较合理的石材图像相似度.
In image retrieval,the probability of HIS sub-space was often used as its features to define the distance and similarity.But these can lead incorrect results in image retrieval.In stone retrieval,for one stone image we must dispatch it into one class and then give its similarity.In this paper,we present an algorithm for stone image retrieval,in which the similarity to one class was given,and then the similarity was computed to each image in the class.At last we get the integral similarity for the stone image.
出处
《武汉理工大学学报（交通科学与工程版）》
2009年第4期768-771,共4页
Journal of Wuhan University of Technology（Transportation Science & Engineering）
基金
国家自然科学基金项目(批准号:50578128)
福州市科技计划项目(批准号:2007-Z-57)资助
