中文摘要：

降雨天气往往导致监控视频质量下降.本文提出首先在对数图像处理（Logarithmic image processing,LIP）框架下利用灰色调约束检测出候选雨滴,进而利用主成分分析（Principal component analysis,PCA）方法计算每个候选雨滴的倾斜方向并构建其概率密度分布函数,利用Mean-shift算法估计该分布函数的峰值,作为检测到的雨滴降落方向,然后,通过方向约束去除候选雨滴中的干扰噪声.最后,文章采用一种加权的重构方法消除雨滴.实验证明,算法能够有效检测并去除各种场景中的雨滴.

英文摘要：

Rain may deteriorate the quality of surveillance videos.In this paper,we propose firstly to detect candidate raindrops in the logarithmic image processing（LIP） framework using constraints of gray tone.Then the tilt angle of each candidate raindrop is calculated using principal component analysis（PCA） and a probability density function of orientation is constructed.Mean-shift is used to estimate the peak of the distribution and it is considered to be the direction of detected falling raindrops,and noise blobs are eliminated by restricting the direction of candidate raindrops.Finally,a weighted composing method is used to remove rain.Experiments have demonstrated that the algorithm is able to detect and remove rain in a variety of scenes.