Disclaimer: these examples are mostly taken from my pre-Eikomagos career
Nieuwenhuizen et al.  published a method to estimate the resolution in a super-resolution image. From the paper the resolution depends on localization uncertainty, density of single fluorescent labels and on the sample’s spatial structure. Their approach estimates the resolution directly from the final image or localisation coordinates.
The authors released their reference implementation in DIPimage. In addition, their group released a plugin for ImageJ/Fiji. However, the latter only implemented a stripped-down version of the methodology. In 2017 Eikomagos was commissioned to improve the plugin. In the end, we started from scratch so both 2D and 3D data sets could be supported using the JTransforms/JLargeArrays package to provide a Fourier transform implementation on large multi-dimensional arrays. The rewritten plugin fully implements:
As we did not develop the methodology, this is a good place to focus on the implementation. imageJ is written in Java and we had no prior experience in Java. As the result shows, this was no obstacle. This example also shows that providing a user interface is very much in scope as well. As we have the scientific background to understand the methodology, we can “think along” and implement it according to your requirements.