Multiscale Intensity Estimation for Marked Poisson Processes

Inference on severely data-starved Poisson processes can be dramatically improved by using auxiliary information about measured discrete events in the form of "marks". Marks are widely available in many applications, and can take the form of photon energy, time delay information, packet size, or other forms of characterizations. Effectively using marks results in innovative signal processing methods and dramatic error reductions for Poisson intensity estimation. The efficacy of the proposed method is demonstrated in the context of photon-limited spatio-spectral intensity estimation.