A unified scheme for morphological analysis based on attribute and relational representations of mosaic image models is proposed. We consider 4 main types of model representation: functional-attribute (2D feature map), functional-relational (4D relational map), structure-resource-attribute (an area list with resources and attributes), and structure-resource-relational (a graph, which nodes correspond to regions and edges – to relations and both having resource attributes). In this case, the forms of representation of the model are equivalent to each other, in the sense that they contain the same information, there is a one-to-one correspondence between them, and the formulas for the transition from one representation to another can be written out explicitly. In this scheme, the construction of specific morphological operator for some complete image model presumes the separation of this model into two parts: the guiding (modifying) part, which determines the transformation algorithm, and the guided (modifiable) part to be transformed. These two parts of the model can intersect, therefore cannot be called “variable” and “constant” components. As a basic sample, we consider the halftone Pyt’ev morphology. We explore the specifics of different-sort models, introduce the mutual models and propose different tools for creation of model-based morphological operators. Further, various other morphological systems can be described and explored using the proposed generalized approach.