中文摘要：

提出一种基于概率距离及融合镜头时空特征的镜头相似性度量方法。横向方面,构造镜头加权颜色直方图来表示时间信息并使用直方图交计算镜头时间相似性;纵向方面,通过非线性映射将空间特征向量映射到高维空间进行高斯分布建模,计算两镜头空间特征高斯分布间的概率距离来度量两镜头空间相似性;最后,对时、空相似性加权融合计算两镜头总相似性。实验结果证明了该方法的有效性。

英文摘要：

This paper proposed a novel method of shot similarity measure based on probabilistic distance and fusing temporal spatial features.To extract the temporal feature,constructed shot weight color histogram and calculated the temporal-similarity by histogram intersection.To calculate the spatial-similarity,at first,mapped the spatial features to a high dimension space by nonlinear mapping,in which they obeyed Gaussian distribution.Then computed the probabilistic distance of spatial features in high dimension space,computed the spatial-similarity by weighting spatial feature probabilistic distance.At last,calculated the shot similarity by weighting temporal similarity and spatial similarity.Theoretical analysis and experimental results show the validity of the proposed method.