In previous works, we investigated the use of local filters based on partialdifferential equations (PDE) to denoise one-dimensional signals through theimage processing of time-frequency representations, such as the spectrogram. Inthis image denoising algorithms, the particularity of the image was hardlytaken into account. We turn, in this paper, to study the performance ofnon-local filters, like Neighborhood or Yaroslavsky filters, in the sameproblem. We show that, for certain iterative schemes involving the Neighborhoodfilter, the computational time is drastically reduced with respect toYaroslavsky or nonlinear PDE based filters, while the outputs of the filteringprocesses are similar. This is heuristically justified by the connectionbetween the (fast) Neighborhood filter applied to a spectrogram and thecorresponding Nonlocal Means filter (accurate) applied to the Wigner-Villedistribution of the signal. This correspondence holds only for time-frequencyrepresentations of one-dimensional signals, not to usual images, and in thissense the particularity of the image is exploited. We compare though a seriesof experiments on synthetic and biomedical signals the performance of local andnon-local filters.
展开▼
