Share Email Print

Proceedings Paper

Neural net architectures for image demosaicing
Author(s): Rhys Buggy; Marco Forte; François Pitié
Format Member Price Non-Member Price
PDF $17.00 $21.00
cover GOOD NEWS! Your organization subscribes to the SPIE Digital Library. You may be able to download this paper for free. Check Access

Paper Abstract

Demosaicing remains a critical component of modern digital image processing, with a direct impact on image quality. Conventional demosaicing methods yield relatively poor results especially light-weight methods used for fast processing. Alternatively, recent works utilizing Deep Convolutional Neural Nets have significantly improved upon previous methods, increasing both quantitative and perceptual performance. This approach has seen significant reduction of artifacts but there still remains scope for meaningful improvement. To further this research, we investigate the use of alternate architectures and training parameters to reduce incurred errors, especially visually disturbing demosaicing artifacts such as moiré and provide an overview of current methods to better understand their expected performance. Our results show a U-NET style Network to outperform previous methods in quantitative and perceptual error and remain computationally efficient for use in GPU accelerated applications as an end-to-end demosaicing solution.

Paper Details

Date Published: 17 September 2018
PDF: 15 pages
Proc. SPIE 10752, Applications of Digital Image Processing XLI, 1075218 (17 September 2018); doi: 10.1117/12.2322353
Show Author Affiliations
Rhys Buggy, Trinity College Dublin (Ireland)
Marco Forte, Trinity College Dublin (Ireland)
François Pitié, Trinity College Dublin (Ireland)

Published in SPIE Proceedings Vol. 10752:
Applications of Digital Image Processing XLI
Andrew G. Tescher, Editor(s)

© SPIE. Terms of Use
Back to Top