|Universal||0||The universal threshold = sigma * Sqrt( 2 * ln n ). Refered to as Visushrink in the literature and tends to oversmooth the signal. However, it is asymptotically optimal in the minimax sense with a hard threshold and thus improves as the number of samples increases.|
|UniversalMAD||1||The universal threshold = sigmaMAD * Sqrt( 2 * ln n ), with sigma esimated using MAD, or Median Absolute Deviation.|
|Sure||2||The SURE threshold, designed to be a smaller that the Universal threshold, is based on Stein's unbiased risk estimate. This threshold should be used with a soft thresholding policy. Refered to as SureShrink or Rigorous Sure in the literature (MATLAB rigrsure).|
|Hybrid||3||Hybrid threshold, often refered to as SureShrink in the literature. When the signal energy is smaller than the noise energy the Universal threshold is used, otherwise the tighter Sure threshold is used.|
Copyright © 2023 CenterSpace Software, LLC. All rights reserved.