Least-squares estimation of multiple abrupt changes contaminated by multiplicative noise using MCMC

This paper addresses the problem of change point detection in signals corrupted by multiplicative noise. Multiplicative noise has been observed in many signal processing applications. These applications include image processing (speckle) or communication systems (fading channels). This paper focuses on edge detection in synthetic aperture radar (SAR) images contaminated by multiplicative speckle noise. However, the proposed approach could also be used for the segmentation of any multiplicative noise corrupted signals or images. When the signal/noise statistics are known, the change point detection problem can be formulated in a Bayesian framework. However, this approach may be intractable in SAR image processing because of the non-Gaussian multiplicative colored noise. The change point can then be estimated using the simple least-squares (LS) algorithm. The main contributions of this paper are to study the Bayesian and LS detectors for edge detection in speckled SAR images.