Crowded regions of the sky present a significant challenge for stellar photometry. I will discuss how we use difference-imaging to extract the signals from variable stars from the background of constant-brightness objects, and show some examples from the KMTNet microlensing survey of the Galactic Bulge. I will briefly discuss the use of Graphical Processing Units (GPUs) for parallel computations in astrophysics, and describe pyDIA, a new GPU-based code for wide-field difference-imaging photometry.