Figure 5

Reconstructed images after the first (a,b) and second (c,d) steps of the algorithm, for the 10-ps delay deconvolved data shown in (Fig. 2). All figures are normalized to their own relative maximum value. The second step of the algorithm is used to refine the image by running a new deconvolution process which uses the peak’s position, given by the first step, and employs a different regularization term (total variation) in order to promote smoothness. For an easier comparison, we show the single frames relative to (e), the “N” arm alone, in (f) the “mirrored-Z” arm alone and in (g) both arms.