Instruments and Control
VŠB-TU Ostrava, 6.5.2003
| Annotation of paper No. 39 |
|---|
| Title: | Ideal Choice of an Orthogonal Basis from a Library of Bases for Denoising |
| Author(s): | KUPCZAK, Marek |
| Text: | This contribution describes a method for ideal denoising. This method is based on a choice of an optimal basis from a library of wavelet orthogonal bases. This is the process of minimizing of the cost function of entropy in contribution with shrinkage method. In particular, the chosen shrinkage methods are the estimation of standard deviation ? and of Donoho-Johnstone sure threshold ? for the wavelet coefficients. The entropy is defined by the discrete distribution of squared wavelet coefficients. The Donoho threshold belongs to the simplest non-linear wavelet shrinkage, concretely to the soft thresholding. The Donoho threshold is more “pessimistic” and more reliable than the hard threshold. So, it is constructed for the “worse signal”. |