
UW inventors and industry collaborators have developed a method to create a low noise video from a noisy signal by combining it with simultaneously captured video of the same scene from a low noise reference camera. In one embodiment a fluorescence signal is very noisy and captured by a fluorescence camera. A second camera simultaneously captures a video of the same sample from the same viewpoint under ambient light (e.g., white light illumination). The video from the second camera does not contain the fluorescence signal, but contains information on different regions in the sample, the sample geometry, and how the sample is moving. The new algorithm uses this information to de-noise the fluorescence video.