Non-linear time-scale contrast for abrupt change detection in multiplicative noise

Multiplicative abrupt changes (AC) have been considered in many applications. These applications include speckle noise in radar images and random communication models. This work applies the continuous wavelet transform (CWT) and possibly nonlinear preprocessing to AC detection in multiplicative noise. CWT translation invariance allows an AC signature to be defined. Then, the problem is signature detection in the time-scale domain. A second-order contrast criterion is defined as a measure of detector performance. The influence of nonlinear pre-processing on time-scale contrast is studied. The performance of two suboptimal signature-based detectors are clearly improved by a nonlinear pre-processing under appropriate conditions regarding noise and signal parameters