In this paper, we present a very general and powerful approach to represent and to visualize the similarity between the objects that contain heterogeneous, imperfect and missing attributes in order to easily achieve efficient analysis and retrieval of information by organizing and gathering these objects into meaningful groups. Our method is essentially based on possibility theory to estimate the similarity and on the spatial, the graphical, and the clustering-based representational models to visualize and represent its structure. Our approach will be applied to a real digestive image database. Without any a priori medical knowledge concerning the key attributes of the pathologies, and without any complicated preprocessing of the imperfect data, results show that we are capable to visualize and to organize the different categories of the digestive pathologies. These results were validated by the doctor.