Indirect 4D nonlocal means
Paulo González, Irène Buvat, Clovis Tauber, Sophie Sérrière, Denis Guilloteau · 2014
In this work, we present a new approach to enhance the signal-to-noise ratio of 3D vector-valued images. We extend the original nonlocal means proposed by Buades to 3D vector-valued images. In the proposed approach, the similarity between vector-valued voxels is calculated indirectly from a smoothed image, avoiding the use of patches and making use of the entire spectral information. Moreover, we introduce weights in the calculus of the similarity to favor low-noise channels. The weights are estimated automatically from a wavelet analysis of each channel. Results on real PET Dynamic acquisitions and GATE Monte Carlo simulations illustrate the potential of the proposed method, which led to distinct improvements of figures of merit over several other approaches from the literature.