基于刚体模型的视频自适应中值去噪算法

传统的去噪研究主要针对图像进行滤波，未考虑视频信号的时域与空域相关性。为充分利用视频信号的时空联系进行去噪，首先建立了视频刚体模型，然后基于刚体模型进行了自适应中值滤波的研究和对比。刚体模型将视频信号分解为只包含平移和旋转运动的刚体块，以线性关系组织视频像素。基于刚体模型的中值滤波使用被滤波点在刚体内空域相邻和时域对应的像素参与中值计算，并根据时域和空域的污染程度不同选择参考像素集的范围。通过对常用测试视频的去噪实验，证实了基于刚体模型去噪的可行性；基于刚体模型的自适应中值滤波的去噪效果优于普通空域自适应滤波，证明了时空联合去噪的优越性。

Video denoising by adaptive median filter based on rigid model

The conventional denoising research mainly focuses on image filters,which can＇t relate the special relativity with the temporal relativity of videos.To take flail advantage of spatial-temporal relations in denoising, A rigid model is established and the adaptive median filtering based on it is researched.The rigid model resolves videos into rigid block containing only parallel and whirl move,organizes the pixels with linear relations.Rigid Model Based Adaptive Median Filter（RBAMF） selects pixels neighbor to filtered pixels in both space and time domain in the rigid for median computation.The scale of referenced pixels is decided by degree of contamination.Denoising experiment is performed with familiar test videos.Experiment Results prove that denoising based on rigid model is feasible.The denoising effect of RBAMF overmatches the effect of special adaptive median filter,which illustrates superiority of spatial-temporal denoising method.