Luciferase reporters have become standard genetic tools to monitor gene expression in real time and in high-throughput using microplate readers. Compared to reporter gene assays based on fluorescence proteins, luciferase reporters have a superior signal-to-noise ratio, since they do not suffer from the high autofluorescence background of the bacterial cell. However, at the same time luciferase reporters have the drawback of constant light emission, which leads to undesired cross-talk between neighboring wells on a microplate. To overcome this limitation, we developed a computational method to correct for luminescence bleed-through and to estimate the "true" luminescence activity for each well of a microplate. As the sole input our algorithm uses the signals measured from a calibration plate, in which the light emitted from a single luminescent well serves as an estimate for the "light-spread function". We show that this light-spread function can be used to deconvolve any other measurement obtained under the same technical conditions. Our analysis demonstrates that the correction preserves low-level signals close to the background and shows that it is universally applicable to different kinds of microplate readers and plate types.