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：
利用了Mean蝴聚类方法对图像进行不规则的划分，在此基础上定义了一种动态局部直方图。设计了一个能同时表征图像颜色和位置信息的统计量，并且给出了计算图像相似度的方法．该算法既改进了以往局部直方图利用图像位置信息的检索方法的不足，而且保留了全局直方图方法具有的旋转不变性和缩放不变性．实验结果表明，在准确引入了图像的空间位置信息后，较大的提高了图像检索的精度，检索结果中不再出现和示例图像颜色成分相似而空间分布不同的图像，实验结果较好的验证了算法的有效性和稳健性．
英文摘要：
The color histogram based image retrieval method doesn＇ t use the spatial distribution information of the image; Image regular segmentation or extracting some main objects, then compute local color histogram will arouse error. In this paper, image is irregularly segmented by Mean shift clustering. Based on image segmentation, one dynamic local histogram is defined, one statistic which can denote image color and spatial distribution information is designed, one image similarity computing method is also represented. This algorithm is better then the local histogram method which using the image location information, the rotation inflexibility and scale inflexibility of the global color histogram method is also reserved. The experiment result show that using image spatial distribution information improve the accuracy of image retrieval, those images color content are similar but the spatial distribution are different will not appear, this algorithm is efficient and robust.
同期刊论文项目
同项目期刊论文
期刊信息
位置：