SAR Image Change Detection Based on Generalized Gaussian Distribution MRF Model

the detection of synthetic aperture radar (SAR) images is getting more and more concern because of the all-weather, all-day geographic observation's characteristic. However, the inherent speckle noise in synthetic aperture radar images leads to a large number of false alarms and missed alarms are generated. Aim at this core problem, this paper proposes to consider the feature information of the neighborhood around the image pixel, and will be able to use the Markov random field model of the local statistical feature of the image to introduce the SAR image change detection by using the dependence of the state between the adjacent coordinates of the image. In the case of Gaussian distribution, it is often impossible to accurately fit the intensity distribution of some pixel points in the changed region. This paper proposes to adopt the generalized Gaussian distribution which has a better fitting effect. The experimental results show that the proposed generalized Gaussian distribution MRF model can better fit the different images' pixel distribution and improve the detection accuracy of the changed region.