Computer vision is nowadays commonly used in automated product inspection. And one common situation arises when the objects present in an input image are in reciprocal contact. The complexity in the shape and color of the objects affects the process of segmentation, whose primary goal does not consist of identifying uniform regions inside the input image, but separating whole objects. A general approach to this situation clearly requires the a prior knowledge of the possible set of objects to be identified and extracted from the input image. The segmentation algorithm must be able to identify the objects before feature extraction and classification.