快速搜索密度峰值聚类在图像检索中的应用

Application of fast search density peak clustering in image retrieval

为缩减图像检索和匹配范围,提高检索速度和准确率,将快速搜索密度峰值聚类用于对图像,按照特征相似性原则进行聚类,在类中心和最接近的一类中进行图像检索。考虑到传统的图像特征提取算法忽略了图像颜色的空间分布信息,提取的特征信息无法突出感兴趣的图像区域,通过等面积矩形环对图像进行划分并计算各空间区域的相关性,根据空间区域相关性计算各区域的重要性,将空间信息与颜色信息进行融合;对聚类算法的截断距离进行合理改进,保证了聚类的精度;将该密度峰值聚类算法应用于图像检索之中。对比实验结果表明,所提聚类算法和空间特征提取方法提高了图像检索的效率和准确性。

To reduce the image retrieval and matching range and improve the retrieval speed and accuracy,the fast search density peak clustering（DP）was adopted to cluster the image according to the feature similarity principle.Image retrieval was executed in range of the class centers and the closest one.Considering that the spatial distribution information of image color is ignored by traditional image feature extraction algorithm,and that the extracted feature information does not highlight the interesting image region,each image was partitioned by equal area rectangle ring and the correlation of the spatial regions was calculated.The importance of each region was calculated according to the spatial correlation to combine the spatial information and color information.The cutoff distance of the clustering algorithm was improved reasonably to ensure the accuracy of the clustering algorithm.The presented density peak clustering algorithm was applied to image retrieval.Experimental results show that the proposed method is feasible and effective.Experimental results show that the proposed clustering algorithm and spatial feature extraction method improve the efficiency and accuracy of image retrieval.