【摘要】 为了进一步提高图像检索系统的效率,提出了一种由谱聚类算法预处理的基于内容的图像检索方法。主要思想是在用户提出检索请求之前,利用谱聚类算法将初始图像库划分为若干子库;用户给出示例图片后,首先搜索与示例图片相似度最大的子库,再将搜索范围锁定在该子库内检索出与示例图片最相似的图片。实验表明,该方法在保持了较高查准率的同时,明显缩短了检索时间,提高了检索效率。更多还原

【Abstract】 A content-based image retrieval method with the preprocessing of spectral clustering is proposed aiming to improve the efficiency of image retrieval systems.The main idea is to partition the initial image set into several subsets before the user’s retrieval request proposes.When the sample image is provided,the method searches for the sample image’s most similar subset first,and then searches for the most similar images only in this subset.Experiment results indicate that while holding the high accuracy,this method shortens the time for retrieval and approves efficiency considerably.更多还原