【摘要】 提出了一种新的聚类方法来支持基于图像或镜头例子的检索 .这种方法以最近特征线 (Nearest FeatureL ine,NFL)聚类方法为基础 ,并根据最近特征线方法的特征 ,将基于特征空间拐点的关键帧提取过程与聚类方法作为一个整体统一考虑 ,从而使得最近特征线方法性能达到最优 .实验结果表明 ,我们的基于关键帧提取的最近特征线方法与传统的最近特征线方法、最近邻法以及最近中心法相比较 ,可以获得更高的检索精度更多还原

【Abstract】 Query by image or video examples is a convenient and effective way to search in video database. This paper proposes a new scheme to support such searches. The main contribution of the proposed scheme lies in considering both the feature extraction and distance computation in feature space together. The distance definition in this paper is a new metric named as Nearest Feature Line (NFL), and the feature to represent a video shot is key frames. So the break point based key frames extraction is combined with the NFL method to achieve a better performance. Experimental results have shown that the combined method achieves superior performance than the traditional NFL, and other classification methods such as Nearest Neighbor (NN) and Nearest Center (NC).更多还原