PLATELETS FOR MULTISCALE ANALYSIS IN PHOTON-LIMITED IMAGING Rebecca Willett and Robert Nowak, ECE Department, Rice University Approximation Rates Wedgelet Partitions Photon-Limited Imaging Photon-limited imaging encompasses applications such as PET, SPECT, confocal microscopy, astronomical imaging, and IR imaging. New alternatives to traditional wavelet analysis are required to tackle C Haar Wavelet Partition Wedgelet Partition Poissonian Observations Image Curves Inverse Problems Beyond Wavelets 2 C Original Image MPLE Algorithm 2 Error decay rates: -1/2) Fourier: O(m -1 Wavelets: O(m ) -1 Wedgelets: O(m ) -2 Platelets: O(m ) Data Const Estimate Wedge Estimate Platelet Estimate Maximum Penalized Likelihood Estimation Platelet Approximations Goal: Maximize the penalized likelihood Wedged Platelet Estimate Inherit from finer scale Nuclear Medicine 5 4.5 4 3 2.5 2 1.5 1 Platelets Wedgelets 0.5 50 5 4.5 4.5 4 4 3.5 3.5 3 3 2.5 2.5 2 2 1.5 1.5 1 1 0.5 0.5 100 150 Terms = 1164 200 150 200 250 Terms = 2068 5 50 100 250 is the likelihood is a smoothing parameter is the number of parameters in So the MPLE is Confocal Microscopy 2.6 MLE (min error level shown in dots) Piecewise Constant Platelet 2.4 2.2 50 100 150 200 250 2 Terms = 510 Error Platelets are localized atoms at various scales, locations, and orientations that can produce highly accurate, piecewise linear approximations to images consisting of smooth regions separated by smooth boundaries. Platelets generalize Donoho’s wedgelets and are capable of concisely representing both smooth surfaces and edges. Haar 3.5 1.8 1.6 Supported by NSF, ARO, and ONR http://www.ece.rice.edu/~willett/ Estimate depends on partition and parameters. 1.4 1.2 2 4 6 8 Iteration 10 12 14