Multi-photon laser scanning microscopy is a cutting-edge tool for high-resolution imaging of living tissue and organisms used to study biological processes such as cellular dynamics. The result of this powerful technique is a series of photons emissions resulting from fluorescence events occurring at each raster scan location. Each observed photon is detected along with auxiliary information such as its characteristic emission spectrum. The small number of photons collected, however, can limit the quality of the reconstructions. This paper proposes a novel reconstruction method which effectively uses the auxiliary information available in multi-photon microscopy to improve the overall reconstruction quality. The multiscale analysis framework underlying the proposed estimator results in reconstructions with accurate estimates of intensities without smoothed edges or boundaries. Unlike conventional multi-dimensional estimators based on tensor products of one-dimensional basis functions...