摘要
由于海量图像中的数量规模大,造成有相同图像检索特征的干扰图像过多,图像低层视觉特征和高层区分特征之间存在的差异度过小,传统的图像检索方法难以充分表达图像的可检索的差异化特征信息,造成图像检索方法鲁棒性不强,检索误差大。提出一种运用图像细微特征区分概念的低复杂度海量图像检索优化方法。针对检索图像的灰度差异,进行图像归一化处理,最大化减小海量图像相似特征对检索方向的消极影响,依据SIFT算法,通过寻求图像极值点的坐标、尺度空间及方向实现对图像细微特征的提取,根据图像颜色的直方图相交的相关理论,计算图像的特征相似度,实现检索图像与目标图像特征相似度匹配,完成对细微特征区分的海量图像检索。实验结果表明,利用改进算法进行细微特征区分的海量图像检索,能够提高检索精度。
A massive image retrieval method with low complexity is proposed by using the concept of image＇s subtle feature distinguish. For the gray level difference of retrieval images,image normalization processing is made,the negative influence of similar features of massive images on retrieve direction is reduced. Based on the SIFT algorithm,by seeking the coordinate,scale space and direction of image＇s extreme point,the extraction of image＇s subtle feature is realized. Based on the related theory of the histogram intersection of image＇s color,the similarity of image feature is calculated to realize the matching of retrieval image with target image feature similarity,and the massive image retrieval is completed in which the subtle feature is distinguished. Experimental results show that the improved algorithm can improve the retrieval precision.
出处
《计算机仿真》
CSCD
北大核心
2015年第3期294-297,共4页
Computer Simulation
基金
江苏省高等职业院校国内高级访问学者计划资助项目(2014FX024
2.2014)
江苏经贸职业技术学院重大课题资助项目(JSJM1401)
关键词
细微特征区分
图像检索
图像内容识别
Subtle feature distinguish
Image retrieval
Image content recognition
