论述了两种基于时域的视频去噪算法。首先介绍基于菱形运动估计的时域滤波，先进行菱形运动估计计算寻找最佳匹配块，之后在两个块之间进行均值滤波，整合匹配块输出滤波过的图像；其次介绍了BM3D视频去噪技术，详细地描述了BM3D去噪技术的去噪主要过程，分别解释了去噪过程中块匹配、三维变换域滤波、权值计算、重构四个主要步骤的原理与过程。最后通过实验，证明了这两种方法在去噪方面均能取得良好的效果。%Two noise elimination algorithms of videos based on time domain are discussed. The time domain filtering based on diamond motion estimation is introduced. The diamond motion estimation is conducted to search for the optimal matching block, and between the two blocks is proceeded with mean filtering. The filtered image is output after matching blocks integration. The BM3D noise elimination technology of videos is presented,and the main process of this technology is described in detail. The principles and process of block matching,three⁃dimensional transformation domain filtering,weight calculation and reconstruc⁃tion in noise elimination process are explained respectively. The experimental results verify that the two methods can achieve good effect in the aspect of noise elimination.
展开▼
