中文摘要：

针对感知噪声导致安防监控视频质量出现退化的问题,提出一种视频感知噪声盲检测算法。通过图像的粗糙集表示,采用邻域粗糙集的相关理论获取视频静态区域的下近似集合,去除运动前景的干扰;引入自适应惩罚机制克服过亮过暗区域对感知噪声产生的掩盖效应,确定最优的有效检测区域,结合信息熵作为该区域感知噪声程度的度量指标,实现对视频感知噪声的盲检测。实验结果表明,该算法在视频存在运动前景和局部含有过亮过暗区域时给出了精确的检测结果,具有良好的稳健性,满足安防场景监控视频的实际需求。

英文摘要：

To deal with the problems that perceptual noise degrades the security surveillance video＇s quality, a blind perceptual noise detection algorithm for surveillance video was proposed. The interference of the motion foreground was removed using the acquired background area＇s domain set based on rough set and neighborhood rough set. An adaptive penalty mechanism was introduced to overcome the masking effects of the-over-bright and the-over-dark areas on the perceptual noise to determine the optimal effective detection regions. The information of regions was taken as the detection index of perceptual noise, realizing the blind detection on monitoring video perceptual noise. Experimental results show that, the algorithm not only has better detection performance on video image with the existence of moving foreground, but also gives encouraging detection results in case of the local image containing the-over-bright and the-over-dark areas. Besides, the consistency between the predict recognition and subjective recognition can meet the actual demand of surveillance video in security and protection scene.