Image processing in various domains like geospatial imaging, astronomy, neuroscience, etc. has seen the size of collected image datasets grow both due to novel techniques to expand the field of view and improve resolution. This presents a number of different challenges to workflows that were often traditionally designed to work in memory. We explore useful primitives in dask and explain how these can be used to aid analyis. Using these primitives, we show users how common image processing functions can be built in this framework and point out their implementations in the dask-image library.