We investigate the improvements in Shack–Hartmann wavefront sensor image processing that can be realized using total variation minimization techniques to remove noise from these images. We perform Monte Carlo simulation to demonstrate that at certain signal-to-noise levels, sensitivity improvements of up to one astronomical magnitude can be realized. We also present on-sky measurements taken with the CANARY adaptive optics system that demonstrate an improvement in performance when this technique is employed, and show that this algorithm can be implemented in a real-time control system. We conclude that total variation minimization can lead to improvements in sensitivity of up to one astronomical magnitude when used with adaptive optics systems.