Image segmentation through using the evidence theory based data fusion technique
A Dromigny-Badin, Yuemin Zhu, Gerard de Mas Giménez, Robert Goutte · 2002
An image segmentation method is presented that is based on a pixel-level data fusion technique employing the evidence theory of Dempster-Shafer (DS). A probability mass being defined for each gray level, each couple of pixels is combined through Dempster's combination rule and a look-up fusion table. The segmentation procedure is iterative, and the determination of probability mass is automatic. The proposed method is illustrated with the aid of both simulations and examples on physical images. The obtained results show the interest of exploiting multiple information for image segmentation.