Fig. 5.

Qualitative deblurring performance of deep learning in comparison with that using blind deconvolution. Two examples of mouse ear vascular images are presented, shown in the entire OR-PAM images (a) and (f). For each example, two ROIs marked with white dashed box are enlarged and compared. In the first example (a), (b1-e1) are for ROI 1 and (b2-e2) for ROI 2; in the second example (f), (g1-j1) are for ROI 1 and (g2-j2) for ROI 2; all correspond to AR-PAM, blind deconvolution, GAN output, and ground truth, respectively. Cross-sectional profiles along the white dashed lines inside (b1-e1), (b2-e3), (g1-j1) and (g2-j2) are provided for comparison. Metrics like PSNR and SSIM with respect to the entire OR-PAM image are also provided for reference. Blind Deconv, Blind Deconvolution.