Video Denoising via Dynamic Video Layering
Han Guo, Namrata Vaswani · IEEE Signal Processing Letters · 2018
Video denoising refers to the problem of removing “noise” from a video sequence. Here, the term “noise” is used in a broad sense to refer to any corruption or outlier or interference that is not the quantity of interest. We develop a novel solution framework, which we call layering denoising (LD), for denoising highly noisy or otherwise corrupted videos that are well modeled as the sum of a low-rank matrix plus a sparse matrix. We show that the performance of existing state-of-the-art denoisers can be significantly improved (especially in large noise settings) if the video is first decomposed into the two layers and the denoiser is applied on each layer separately. Our proposed solution uses a recursive projected compressive sensing (ReProCS) based algorithm for the layering task and video block matching and three-dimensional filtering for denoising each layer. We show the power of our proposed approach, ReProCS-LD, using exhaustive experimental comparisons.