分层聚类矩形分割的图像/视频编码与分析（英文）

Divisive hierarchical clusteringon cuboids for fast image/video codingand analysis

基于抽象区域的图像检索(RBIR)对于查询相关图像是有效的,它将构成智能视频监控的基础,其中感兴趣区域(ROI)的特征是实时提取的。对基于经典聚类方法的现有图像分割技术进行综述,包括一种新颖的矩形分割(CSeg)技术,该技术生成近似的矩形图像片段。CSeg具有最小可能的线性顺序运行时间,在性能方面,使用近似段的RBIR在统计上类似于那些使用不适于实时应用的精确分割算法的RBIR。

Region-based image retrieval （RBIR） has been proven to be effective in finding relevant images. It will form the basis for intelligent video surveillance where features of regions-of-interest （ROI） is extracted in real-time. This paper presents a review on existing image segmentation techniques based on the classical clustering approaches, including a novel Cuboid Segmentation （CSeg） technique that results in approximated rectangular image segments. CSeg is fast with minimum possible linear order running time and performance of RBIR using the approximated segments is statistically similar to RBIR using accurate segments from segmentation techniques that are infeasible for real applications.