Fig. 4 (a) is a ground-truth object, a, and (b) is a data image, d. It is a, convolved with a ST ScI-supplied PSF, p, and corrupted with Poisson noise. The max and min of d are 14,847 and 23. Fig. 4(c) shows a Wiener filter estimate, â, with z = 0.005. Fig. 4(d) shows the CR bound on the estimate. It was calculated based on the estimate, â, not on the ground-truth image, a.
Fig. 5 shows two 1-D cuts through the images and bounds. On the left are the sample RMSE's for 50 trials of the estimate. On the right is the CR bound for the first trial, based on the first estimate. We see that the bound does indeed provide a reasonable estimate for the MSE associated with the estimate.
