Fast 3D non-local denoising algorithm for video image using inter-frames dependency

We presented a novel 3D non-local denoising algorithm for video enhancement. The main advantage of our algorithms is exploiting the correlation of patches in multi-frames. Meanwhile, assuming existence of larger similarity, we take advantage of temporal and spatial redundancies to estimate filter rather than calculate Euclidean distance. Comparative results show that our algorithms can be about 3 times faster than traditional non-local algorithms, while achieving almost equal denoising performance.