Disjunctive fusion does not reach any consensus between the sources since it assumes them unreliable. On the contrary, it puts them in competition. Reaching a consensus discards the erroneous values due to noise. By putting the sources in competition, disjunctive fusion favours these errors.

However, it should be recognized that the problem of unclassified pixels is solved by disjunctive fusion. Thus, even if the number of sources to be fused becomes large, large areas of unclassified pixels do not appear. In effect, a pixel is classified if one among all the sources gives a degree of membership not null to any of the classes.

Another problem arises: the impossibility of choosing the class to which a pixel must be assigned, when several classes are possible with the same maximum degree.

The disjunctive fusion interests are the following:

• an easy implementation;
• a linear complexity;
• fast running according to linear complexity property and operator ``maximum'' which has a low cost;
• the property of associativity is satisfied.
  
  

But the disadvantages of such an approach are important:

• an insensibility to the conflict between sources;
• an indecision about classifying a pixel, particularly due to the artificial disappearance of conflict between the sources ;
• a behaviour too exclusive with sources as they are supposed not in agreement, except one;
• an inefficiency for the fusion of images due to a behaviour too exclusive.