State-of-the-art modelling algorithms allowing for prediction of macroscopic properties of martensitic steels are primarily based on microstructural parameters such as prior austenite grain size as well as martensite packet-and block size distribution. The latter are usually obtainable via electron backscatter diffraction (EBSD) measurements. However, determination via light microscopy would present a more cost-effective determination method. This work presents a python-based multi-step image processing algorithm capable of separating the grain boundaries and the martensitic substructure from etched micrographs. Additionally, the viability of a characteristic mean free path parameter λ of the martensitic substructure for comparison of different martensitic microstructures is tested. To this end, a microstructure variation of PH15-5 was performed using different heat treatments and the specimens were analyzed using EBSD and electrolytically etched micrographs.