Current browse context:
eess
Change to browse by:
References & Citations
Computer Science > Computer Vision and Pattern Recognition
Title: Deep Fusion Prior for Plenoptic Super-Resolution All-in-Focus Imaging
(Submitted on 12 Oct 2021 (v1), last revised 15 Oct 2022 (this version, v5))
Abstract: Multi-focus image fusion (MFIF) and super-resolution (SR) are the inverse problem of imaging model, purposes of MFIF and SR are obtaining all-in-focus and high-resolution 2D mapping of targets. Though various MFIF and SR methods have been designed; almost all the them deal with MFIF and SR separately. This paper unifies MFIF and SR problems in the physical perspective as the multi-focus image super resolution fusion (MFISRF), and we propose a novel unified dataset-free unsupervised framework named deep fusion prior (DFP) based-on deep image prior (DIP) to address such MFISRF with single model. Experiments have proved that our proposed DFP approaches or even outperforms those state-of-art MFIF and SR method combinations. To our best knowledge, our proposed work is a dataset-free unsupervised method to simultaneously implement the multi-focus fusion and super-resolution task for the first time. Additionally, DFP is a general framework, thus its networks and focus measurement tactics can be continuously updated to further improve the MFISRF performance. DFP codes are open source available at this http URL
Submission history
From: Yuanjie Gu [view email][v1] Tue, 12 Oct 2021 02:44:07 GMT (2191kb)
[v2] Sun, 8 May 2022 11:13:57 GMT (1782kb)
[v3] Thu, 19 May 2022 13:29:57 GMT (1772kb)
[v4] Thu, 30 Jun 2022 11:49:13 GMT (1525kb)
[v5] Sat, 15 Oct 2022 11:04:18 GMT (9794kb)
Link back to: arXiv, form interface, contact.