Unsupervised multisensor image segmentation using consonant belief function

In this paper, we shall describe a new iterative method of unsupervised multisensor image segmentation based on the evidence theory. We show that the modeling by means of evidence theory is well suited to the processing of redundant and complementary data as the satellite images. This theory turns out to be quite efficient in unsupervised multisensor image segmentation. The application of the evidence theory in fusing information coming from different sources still poses certain problems. Of paramount importance is the problem of estimating the belief functions. In our work, consonant belief functions were used to represent images. Thus, a technique based on the expectation maximization (EM) algorithm is applied to estimate the consonant belief functions. The EM algorithm has been well known as a convenient and efficient tool to iteratively compute the maximum likelihood estimates. The effectiveness of the proposed method is demonstrated on synthetic and real images.