Fig. 1. Schematic diagram of PhysenNet.

Fig. 2. Experimental flow chart and pseudo code. (a) dx optimization. (b) dx fixed optimization.

Fig. 3. Experimental setup.

Fig. 4. (a) Ground truth. (b) Diffraction pattern (z = 40 mm, pixel_size = 0.0045 mm). (c) Restored image by GS algorithm (iteration = 10000). (d) Restored image by HIO algorithm (iteration = 10000).

Fig. 5. (a1)–(a10) Diffraction patterns corresponding to different diffraction distances z. (b1)–(b10) Recovered images corresponding to different diffraction distances z with trained dx using algorithm A.

Fig. 6. Restored images for different z with different fixed dx values using algorithm B.

Fig. 7. Specific SSIM values for different z with different fixed dx. (a) z = 10–50 mm; (b) z = 60–100 mm; (c) z = 110–150 mm; (d) z = 160–200 mm.

Fig. 8. Specific SSIM values for different z corresponding to different best dx values.

Fig. 9. Restored images with different Δz when the diffraction distance (a) z_measured = 50 mm and (b) z_measured = 100 mm.

Fig. 10. (a) When z_measured = 50 mm, the trained dx (dx_adjust) and SSIM values obtained using different z_input and algorithm A corresponding to the restored images in Fig. 9(a). (b) When z_measured = 100 mm, the trained dx (dx_adjust) and SSIM values obtained using different z_input and algorithm A corresponding to the restored images in Fig. 9(b). (c) Tolerance of the distance uncertainty. Error bars denote the maximum tolerance of the initial value of z_input with respect to z_measured that PhysenNet is allowed.