Skip to main content
. Author manuscript; available in PMC: 2023 Apr 1.
Published in final edited form as: IEEE Trans Pattern Anal Mach Intell. 2022 Mar 4;44(4):1934–1948. doi: 10.1109/TPAMI.2020.3033882

TABLE 4. Comparison of FlatNet with Tikh+U-Net.

The top half compares FlatNet-sep with Tikh+U-Net for separable lensless model while the bottom half compares FlatNet-gen with the corresponding Tikh+U-Net. FlatNet outperforms Tikh+U-Net for both separable and non-separable models because it learns an end-to-end mapping.

Methods PSNR (in dB) LPIPS
Separable Model

Tikh+U-Net 18.90 0.322
FlatNet 19.62 0.256

Non-separable Model

Tikh+U-Net 20.60 0.298
FlatNet 20.94 0.296