RT journal article
T1 The dual wavelet spectra: an alternative perspective on Hurst exponent estimation with application to mammogram classification
A1 J. Hinton, Raymond
A1 Vidakovic, Brani
A1 Ramírez Cobo, Josefa
A2 Estadística e Investigación Operativa
K1 Scaling
K1 non-decimated wavelet transform
K1 2D fractional Brownian motion
K1 image classification
AB The wavelet spectra is a common starting point for estimating the Hurst exponentof a self-similar signal using wavelet-based techniques. The decay of the log2 averageenergyofthedetailwaveletcoefficientsasafunctionofthelevelofsignaldecompositioncan be used to construct estimators for this parameter. In this paper, we expandon previous work which introduced the “dual” wavelet spectra, where decompositionlevels are instead treated as a function of energy values, and propose a relationshipbetween its slope and the Hurst exponent, creating a new estimator. We focus onthe 2D case, making the method suitable for analyzing complex image data. Theeffectiveness of this estimator and its sensitivity to several settings are demonstratedthrough a simulation study. Finally, we show how the technique performs as a featureextraction method by applying it to the task of detecting the presence of breast cancerin mammogram images. Dual spectra wavelet features had a statistically significanteffect on the log-odds of Cancer.
YR 2026
FD 2026-02
LK http://hdl.handle.net/10498/38829
UL http://hdl.handle.net/10498/38829
LA eng
DS Repositorio Institucional de la Universidad de Cádiz
RD 07-nov-2026