Automatic analysis of the difference image for unsupervised change detection

One of the main problems related to unsupervised change detection methods b ased on the "difference image" lies in the lack of efficient automatic tech niques for discriminating between changed and unchanged pixels in the diffe rence image, Such discrimination is usually performed by using empirical st rategies or manual trial-and-error procedures, which affect both the accura cy and the reliability of the change-detection process. To overcome such dr awbacks, in this paper, we propose two automatic techniques (based on the B ayes theory) for the analysis of the difference image. One allows an automa tic selection of the decision threshold that minimizes the overall change d etection error probability under the assumption that pixels in the differen ce image are independent of one another. The other analyzes the difference image by considering the spatial-contextual information included in the nei ghborhood of each pixel. In particular, an approach based on Markov Random Fields (MRF's) that exploits interpixel class dependency contexts is presen ted. Both proposed techniques require the knowledge of the statistical dist ributions of the changed and unchanged pixels in the difference image. To p erform an unsupervised estimation of the statistical terms that characteriz e these distributions, we propose an iterative method based on the Expectat ion-Maximization (EM) algorithm. Experimental results confirm the effective ness of both proposed techniques.