Dempster-Shafer's rule implementation

Dempster-Schafer's Rule Implementation

For any pixel x of an image I_i, we must distribute the mass of belief on all the classes to recognize, to indicate with which degree the information contained in image I_i states that a pixel x is likely to belong to class c.

The mass $m(\theta)$ allotted to uncertainty, denoted $\theta$ , is a tool which allows uncertainty of the source I_i on the choice of the class of membership of the pixel x, to be represented. We can consider that $\theta$ corresponds to any of the k possible classes: $\theta = C_1 \cup C_2 \cup \ldots \cup C_k$ .

The difficulty thus lies in the distribution of the initial mass of evidence between all the classes, and especially in the calculation of the part which must return to uncertainty $\theta$ .

The distribution of the mass between each class consists in giving a proportion of this mass equivalent to the degree of possibility, allotted by the source, to the fact that the pixel x is an element of the class c.

Example

IRIT-UPS