Deep image prior for polarization image demosaicking

The Division of Focal Plane (DoFP) image sensors can obtain the polarization information of the object surface. However, the polarization images generated by the DoFP image sensors have the problem of instantaneous field of view (IFoV) error and resolution reduction. To solve this problem, we propos...

Full description

Saved in:
Bibliographic Details
Published in2023 4th International Conference on Big Data & Artificial Intelligence & Software Engineering (ICBASE) pp. 154 - 157
Main Authors Shi, Yinxia, Wen, Desheng, Jiang, Tuochi
Format Conference Proceeding
LanguageEnglish
Published IEEE 25.08.2023
Subjects
Image sensors
Interpolation
neural network
Neural networks
Polarization
PSNR
Training
Visualization
Online AccessGet full text

Cover

More Information
Summary:The Division of Focal Plane (DoFP) image sensors can obtain the polarization information of the object surface. However, the polarization images generated by the DoFP image sensors have the problem of instantaneous field of view (IFoV) error and resolution reduction. To solve this problem, we propose a method based on deep image prior to solve this problem. Our method employs direct adaptation of the generator network to a single low-resolution polarized image and does not require training on a large dataset. Comparing our proposed method with other polarization image interpolation methods, our method performs better in visual, peak signal-to-noise ratio (PSNR) and structural similarity index (SSIM).
DOI:10.1109/ICBASE59196.2023.10303066