欢迎您！东篱公司
退出
-
申报数据库
-
立项数据库
-
成果数据库
-
项目获奖数据库
中文摘要：
准确计算图像多特征距离成为大数据时代影响基于图像的内容标签的一个关键问题，对基于内容的图像检索技术至关重要。在借鉴欧氏距离和高斯归一化两种方法的优势的基础上，对高斯归一化算法进行改进，提出一种基于特征距离纠偏的多特征距离计算算法。该算法首先采用欧氏距离法计算定量特征距离，然后利用改进高斯归一化法完成距离纠偏，最后通过自由设定权重得到最终的图像多特征距离。与传统高斯归一化算法进行比较，实验结果表明，利用该算法既能有效得到特征间的定量距离，又能方便地把多个特征的地位均衡，从而达到提高相似图像搜索质量的目标。
英文摘要：
The accurate calculation of image’s multi?feature distance is a key problem in big data era,which influences onthe image? based content label,and plays an important role in content?based image retrieval technique. On the basis of the ad?vantages of Gaussian normalization method and Euclidean distance method,the Gaussian normalization method is improved,and a multi?feature distance calculation(C?GN)algorithm based on feature distance rectification is presented. The Euclidean dis?tance method is used in the C?GN algorithm to calculate the quantitative feature distance,and then the improved Gaussian nor?malization method is used to rectify the distance. The image’s multi?feature distance is obtained through the free weight setting.The experimental results show that, the algorithm can not only effectively obtain the quantitative distance among the features,but balance the status of multi?features conveniently,which improve the search quality of similar images.
同期刊论文项目
同项目期刊论文
期刊信息
位置：