Conjunctive fusion has a very severe behaviour on the images. That partly comes from the nature of the data to fuse. Indeed, the process of fusion occurs at a pixel level.

The only data to be fused are thus the degrees of membership of the pixel to each class, provided by each source. For four sources to be fused, four values per pixel for each class are fused, and not four possibility distributions.

The fusion of specific values rather than whole distributions reinforces the severe effect of conjunctive fusion. It is enough that only one among the sources sets a null degree of membership to a pixel to discard a given class from the possible solutions. A class is retained when all the sources attribute a degree of membership not null to this class.

The addition of sources of information to the process of fusion improves the discrimination of the classes, but increases also the number of unclassified pixels. It is indeed more difficult to have ten sources which support all the membership of a pixel to several classes than four sources, and the distinction between the sources is better.

The interest of conjunctive fusion is due to:

• an easy implementation;
• a linear complexity;
• a fast running according to linear complexity property;
• the property of associativity.

On the other hand, it can be reproached to conjunctive fusion:

• a too severe behaviour for the data to fuse;
• a great sensitivity to the conflict between the sources, due to an instability of the behaviour of the operator around zero: the unanimity is needed to take a decision and if only one source cannot decide, then fusion fails;
• an inefficiency for the fusion of a great number of sources, due to the inability of managing the conflict between the sources.