采用背景提取和自适应滤波的视频降噪算法

Video Noise Reduction Method Using Background Extraction and Adaptive Spatial Filter

针对监控视频图像背景固定的特点,提出一种有效去除高斯噪声和脉冲噪声的降噪算法。首先通过分析噪声设计一种提取视频序列背景图像的算法,然后对运动区域采用自适应像素域滤波算法来进行处理。该算法根据最小可觉差和视频图像特征自适应地选择谐波均值滤波、加权算术平均滤波、α-截尾均值滤波和中值滤波。为评估降噪算法性能,将降噪处理前后的视频序列分别进行MPEG-2编码,并改变目标码率对比视频质量。实验结果显示:降噪处理后的视频能够用更少的(约50%)比特数获得相同的主、客观视频质量;或者用相同的比特数获得更高的视频质量。

An effective noise reduction method is proposed for video surveillance image sequences,which have a fixed background and usually corrupted by Gaussian noise or impulse noise.First,an algorithm is designed to extract the constant background,and the current picture is subtracted from the background.Then the object area is processed using an adaptive spatial filter.The adaptive scheme,which combines spatial just noticeable distortion(JND) with local image characteristics,is composed of a harmonic mean filter,a weighted arithmetic mean filter,a alpha-trimmed mean filter,and a median filter.Experimental results of MPEG-2 codec show that the preprocess of noise reduction achieves the same video quality with lower bit rate(about 50%),or high video quality with the same bit rate.