为了去除单帧图像小波去噪后残留的噪声和去噪时引入的类似脉冲噪声的伪细节,提出 一种基于运动补偿的三维KNN (K-Nearest Neighbors)帧间滤波视频序列去噪方法。该方法首先对含噪声的视频序列中的每一帧图像进行小波去噪,然后对去噪后的图像进行基于运动补偿的三维KNN 帧间滤波。实验结果显示,本文中提出的方法可以有效去除视频序列中的噪声,同时可以很好地保持运动对象的边缘。 关键词:小波去噪;视频去噪;帧间滤波;运动补偿 Abstract: A novel wavelet video denoising method is proposed to remove the residual noise and artifacts after wavelet denoising of individual frames. The method processes each frame of video in two steps: (i)denoising each frame in the wavelet domain, and (ii)3D KNN (K-Nearest Neighbors)filtering based on motion compensation. Experimental results show this method removes noise significantly and preserves motion object boundaries. Key words: wavelet denoising, video denoising, inter-frame filtering, motion compensation