监控视频横向条纹扰动快速盲检测算法

A FAST BLIND DETECTION ALGORITHM FOR HORIZONTAL STRIPES DISTURBANCE IN SURVEILLANCE VIDEO

针对视频中可能出现的横向条纹扰动会导致大量信息丢失的问题,提出一种基于时空域特性的快速条纹扰动盲检测算法。该方法首先总结了条纹颜色、形状以及运动性三方面特征。通过利用条纹的颜色和形状特征,计算行累计分布和行方差分布,得到空域内的条纹检测指标。再利用条纹的移动特性,通过帧间差分法和基于对称性的滤波获得条纹滤波图像,进而计算行方向的数据分布,得到时域内的检测指标。最后通过对时空域内检测结果的融合完成对条纹扰动的最终检测。实验结果表明该方法鲁棒性高、速度快、能够准确地检测视频条纹扰动。

To solve the problem of large amount of information losses caused by horizontal stripe disturbance possibly occurred in video,this paper presents a spatiotemporal domain characteristics-based fast blind detection algorithm for stripe disturbance. The algorithm first summarises the characteristics of stripes in three aspects including the colour,the shape and the movability. The first two characteristics are used to calculate the row cumulative distribution and row variance distribution,and gets the stripe detection index in spatial domain. Then,by utilising the third characteristic and through interframe difference method and symmetry-based filtering to get the stripe filtering image and further to calculate the data distribution in row direction,the algorithm obtains the detection index in temporal domain. At last,the final detection on stripes disturbance is completed by integrating the detection results within the spatiotemporal domain. Experimental results show that the proposed method has high robustness and fast speed and can accurately detect video stripes disturbance.