The phase inversion process was investigated from different points of view. First, a reliable on-line and off-line method of morphology detection using light microscopy is established. The method provides fast information that can be used to discriminate between different types of morphologies.Second, TEM images are modelled with Markov Random Fields. For characterization purposes image features are introduced and tested. A combination of features allows classification of different morphologies of PMMA/PS blends.In the third part we show that the rheological properties of blends allow also discrimination between different morphologies.