Markov-based approaches for ternary change detection between two high resolution synthetic aperture sonar tracks

Change detection methods, consisting in detecting potential changes between two images, representing the same geographical area but acquired at different times, have been widely used in sonar imagery. Such methods are very useful to accurately monitor, potentially low, variations in complex environments. Coherent methods, relying on both amplitude and phase of the backscattered signal, are limited because of the signals correlation that plummets with both frequency and temporal baseline. In this paper, we propose an incoherent method, only using the amplitude images, to solve the change detection problem. This method relies on a robust mathematical expression for the class conditional probability density functions of the log-ratio image along with various Markov-based approaches to provide a ternary change map, thus allowing to better understand the changes that have occured on the seafloor.