Image segmentation or image understanding ?

In Image processing, it is often implicitely admitted that an image is known when it has been segmented in various textures, in accordance with the judgement of the human eye. Two counter examples are presented, in biology and in petrology, showing that image analysis may also consist in synthetizing informations disseminated over a large number of pictures, and in predicting the relationships between physical properties and textures. Then, a formal definition of image understanding is given, which yields a particular approach based on structuring elements, and on a few logical properties. An illustration of the approach is provided by the analysis of a boolean structure.