Our software SHERPA enables the automated mass-analysis of diatom shapes by offering a versatile image processing workflow focused on the identification and measurement of object outlines. It handles all steps from image segmentation over object identification to feature extraction with minimal user interaction, extracting of a wide range of outline shape descriptors widely used in diatom studies and elsewhere. Targeting a classical system in polar paleo-oceanography, we analyzed the morphometric variability of Fragilariopsis kerguelensis valves throughout a Southern Ocean sediment core, representing glacial and interglacial periods. Commonly only few features are actually measured for this kind of analysis, whilst e.g. the valve-area is just estimated from the valve’s length and width by applying a correction factor. SHERPA however accurately measures the valve-area, and from these measurements we can identify morphological trends in Fragilariopsis kerguelensis in glacial/interglacial periods more precisely than was previously possible. These analyses also showed that in the last glacial maximum, two clearly distinct morphotypes of F. kerguelensis coexisted. These morphotypes differ in shape-characteristics, thus resulting in different area-correction factors, which are equivalent to the morphometric descriptor “rectangularity” as it is derived by SHERPA from actual measurements. These novel findings in a heavily studied system nicely illustrate the potential gains that can be reached by automated image analyses in diatom morphometrics.