V2V3D: View-to-View Denoised 3D Reconstruction for Light Field Microscopy

Zhenqi Fu, Jiayin Zhao, Tao Yu, Hui Qiao

Abstract:

Light field microscopy (LFM) has gained significant at
tention due to its ability to capture snapshot-based, large
scale 3D fluorescence images. However, existing LFM
reconstruction algorithms are highly sensitive to sensor
noise or require hard-to-get ground-truth annotated data
for training. To address these challenges, this paper in
troduces V2V3D, an unsupervised view2view-based frame
work that establishes a new paradigm for joint optimiza
tion of image denoising and 3D reconstruction in a uni
f
ied architecture. We assume that the LF images are de
rived from a consistent 3D signal, with the noise in each
view being independent. This enables V2V3D to incor
porate the principle of noise2noise for effective denois
ing. To enhance the recovery of high-frequency details, we
propose a novel wave-optics-based feature alignment tech
nique, which transforms the point spread function, used for
forward propagation in wave optics, into convolution ker
nels specifically designed for feature alignment. Moreover,
we introduce an LFM dataset containing LF images and
their corresponding 3D intensity volumes. Extensive exper
iments demonstrate that our approach achieves high com
putational efficiency and outperforms the other state-of-the
art methods. These advancements position V2V3D as a
promising solution for 3D imaging under challenging con
ditions. Our code and dataset will be publicly accessible at
https://joey1998hub.github.io/V2V3D/.