It is noted that the calculation of the degrees of consensus locally to each pixel provides a modification of the behaviour of adaptive fusion, which becomes more sensitive to the data to fuse. In fact, the progressive change from conjunctive fusion to disjunctive fusion disappeared.

With possibility distributions, the process passes gradually from a fusion of type conjunctive towards a fusion of type disjunctive because we define:

• intervals over the distributions, for which conjunctive fusion (area of intersection between the possibility distributions) is applied, and
• other intervals for which disjunctive fusion is used (areas out of the intersection).

However in the case of scalars, there is no notion of intervals to which are applied fusions of different types. It is necessary on the other hand to determine if the involved scalars contribute to find a common plausible class (scalar not null) or not (null scalar):

• if all the sources reach a consensus for a particular class, fusion will be of conjunctive type
• if the unanimity cannot be reached, then fusion is systematically of disjunctive type.

The adaptive behaviour is then lost, and turned into a binary system where conjunctive fusion or disjunctive fusion is chosen depending on whether the unanimity for the classification of the pixel in a class has been obtained or not.

In addition, it should be noted that adaptive fusion is discontinuous around zero. If the possibility distributions have a very weak intersection, this one is used to apply a renormalized conjunctive fusion whereas the agreement of the sources for this area is weak. However, if the distributions are completely distinct, then conjunctive the fusion part is not used. Instead of a progressive transition from the state ``no conjunctive fusion'' to the state ``conjunctive fusion'' occurs a brutal one.

Lastly, the calculation of the height of the intersection between the possibility distributions presents the double disadvantage:

• to have a complexity in . This complexity is due to the cost of the sort in decreasing order of the possibility measurements for each pixel in order to allow the degree of consensus to be calculated.
• to make adaptive fusion operator lose the property of associativity.

• Disadvantages and Advantages