基于压缩域编码长度的视频显著性检测

生物学研究表明,人会 明显地注意 视频中的运动目标。为模拟该特性并快速完成视频显著图的计算,提出一种压缩域时空显著度检测方法(Temporal-Spatial Saliency in Compress Domain model,TS2CD)。分别利用H.264视频中对宏块的残差编码长度和运动矢量编码长度模拟人眼的显著性刺激强度,从而得到视频显著特征。通过线性的加权融合算法,综合两种编码长度得到的空域显著图和时域显著图,得到最终的视频显著图。在3个公开的数据库上的实验表明,TS2CD算法是当前性能最优的方法。

Video Saliency Detection Based on Compressed Domain Coding Length

Biology studies show that people will pay much attention to the moving object when they watch a video.In order to simulate this feature and detect the salient region rapidly,we proposed the temporal-spatial saliency in compress domain model(TS2CD).By respectively using H.264 residual coding length and motion vector coding length,we simulated the salient stimulus intensity and then got video saliency features.Finally,we used the linear weighted fusion algorithm to get the final video saliency maps.Experimental results on three public datasets demonstrate that our model outperforms state-of-the-art methods.