The interpretation of images is at the heart of most investigations in astronomy, and much effort is spent on designing and operating telescopes and instruments that produce the sharpest and finest possible data-products. There are however a few important scenarios where despite these efforts, the information one cares about remains clouded by residual instrumental effects. When looking for instance at AO corrected high-contrast images, one needs to disentangle genuine astrophysical structures surrounding the target from residual diffaction features. The discrimination tool is the fundamental measurement of the degree of coherence, the usual data-product of interferometry. Looking at images as the result of an interferometric process makes it possible to ge beyond the conventional diffraction limit. I will describe one possible operating mode of this point of view that is currently being investigated in the context of a European project called KERNEL and highlight two applications: the post-processing of mid-to-high Strehl images for high-precision relative astrometry of low-mass binaries; a focal plane based wavefront control loop successfully deployed on-sky by the extreme adaptive optics instrument SCExAO at the Subaru Telescope.