A hybrid wavelet and temporal fusion algorithm for film and video denoising

Noise is an impairment which often occurs in both film and digital video and severely degrades the viewing experience of the content. In this work, we propose a two-phase algorithm for film an video denoising. In the first phase, the concept of semi-local shrinkage functions is used to effectively separate noise from image structure. In the second phase, we show how to fuse the result images of the first phase in a way which is robust to motion-compensation errors. A quantitative evaluation of the proposed algorithm on a realistic test dataset with noise of different coarseness and magnitude shows that the proposed method delivers better results than the video denoising method CVBM3D.