TOWER - Telescopic Optimal Wavelet Estimator of the Risk

Removing noise from data is a well studied problem in the mathematical literature. A recent class of methods involves a transform of the data in the wavelet domain and a subsequent shrink of the coefficients by means of a suitable function. We introduce a shrinking function (TOWER) that arises from an estimate of the optimal $L_2$-risk. Univoqueness of the function is shown, when a first-guess solution of the problem is given. Numerical experiments are worked out, when the first-guess is obtained by wavelet regularization (Red-TOWER)