Impulsive Noise Removal from Gray-Scale Video Sequences via Adaptive Thresholding

In this paper, we propose a novel method to reconstruct a video sequence corrupted by impulsive noise. Images and videos can become sparse through transformations such as discrete cosine transform, Wavelet, and contourlet. We can employ this redundancy in the images and videos and the sparsity of the impulsive noise in the observation domain to recover the noisy video sequence. By introducing the augmentation of the vectorization of frames, we convert the video denoising problem to image denoising and employ the temporal information in our algorithm as well. We then compare our method with the baseline 3-D median filter for random-valued impulsive noise removal from a gray-scale video sequence. Simulation results show that our algorithm is simple and very fast, which makes it a suitable choice for online applications. Moreover, its reconstruction quality is far better than that of the classical method, 3-D median filter.