Locating subsurface sources from passive seismic recordings is difficult when attempted with data that have no observable arrivals and/or a low signal-to-noise ratio. Energy can be focused at its source using time-reversal techniques. However, when a focus cannot be matched to a particular event, it can be difficult to distinguish true focusing from artifacts. Artificial focusing can arise from numerous causes, including noise contamination, acquisition geometry, and velocity model effects. We present a method that reduces the ambiguity of the results by creating an estimate of the signal-to-noise ratio in the image domain and defining a statistical confidence threshold for features in the images. To do so, time-reverse imaging techniques are implemented on both recorded data and a noise model. In the data domain, the noise model approximates the energy of local noise sources. After imaging, the result also captures the effects of acquisition geometry and the velocity model. The signal image is then divided by the noise image to produce an estimate of the signal-to-noise ratio. The distribution of image signal-to-noise ratio values due to purely stochastic noise provides a means by which to calculate a confidence threshold. This threshold is used to set the minimum displayed value of images to a statistically significant limit. 2D synthetic examples show the effectiveness of this technique under varying amounts of noise and despite challenging velocity models. Using this method, we collocate anomalous low-frequency energy content, measured over oil reservoirs in Africa and Europe, with the subsurface location of the productive intervals through 2D and 3D implementations.