De-noising algorithm using sparse 3D transform-domain collaborative filtering and adaptive soft thresholding

Video block matching and 3D filtering algorithm for de-noising, which is called V-BM3D is very popular because of its high performance. This paper presents an improved version of V-BM3D in terms of perceptual visual quality. Instead of traditional hard-thresholding of V-BM3D, the proposed algorithm employs adaptive soft-thresholding by applying weak thresholds for lower frequency components and strong thresholds for higher frequency components. So, we can successfully preserve low- and middle-frequency details which mostly affect perceptual visual quality. In addition, we propose a modified weighting strategy for aggregation to mitigate performance degradation due to motion compensation errors. Simulation results prove that the proposed algorithm provides better visual quality than conventional V-BM3D with higher PSNR of 0.3dB at maximum.