摘要
随着数据时代的来临,对互联网上海量的图片数据如何进行更加有效的搜索成为一个热门的问题。关键词搜索存在语义表达问题,而传统的基于内容的图像检索(CBIR)又存在准确率和耗时等问题,由此提出一种新的基于SIFT的图像检索特征改进方法。不同于传统图像检索中使用的直方图,轮廓线等特征方法,实验中利用尺度不变特征变换(SIFT),我们提出了一种新的基于空间特征加权的模糊C均值聚类方法(FWCM)来进行图像特征降维,利用聚类降维后的特征,使用K-D树算法进行多分类预测分析,最终通过多组对比实验,证明了该方法在保证效率的同时,对图像检索的准确率也有着显著的提升。
With the advent of data era, how to search the huge amount of image data on the intemet comes to a hot issue. Key words search has the problem of semantic expression, while the traditional content based image retrieval （CBIR.） has the problem of accuracy rate and time consuming, and so a new method based on the SIFT for improving image retrieval is proposed.Different from the traditional feature extraction methods, such as histogram and contour lines, in our experiment ,we used the scale invariant feature transform （SIFT） and proposed a new method, fuzzy weighted C mean clustering （FWCM） based on spatial features to reduce image features dimension,and then we use the feature simplied by clustering to analyze the multi classification forecast by K-D tree method. Finally, through a multi group contrast experiment, it is proved that the method is efficient, and has a good promotion in the accuracy of image retrieval.
出处
《数字技术与应用》
2016年第1期139-141,共3页
Digital Technology & Application
关键词
尺度不变特征变换
空间特征
加权模糊C均值
多分类预测K-D树
Scale mvariant feature transform Spatial feature Fuzzy weighted C mean clustering Multi classification forecast. K-D tree
