PyNET-Q×Q: An Efficient PyNET Variant for Q×Q Bayer Pattern Demosaicing in CMOS Image Sensors

Deep learning-based image signal processor (ISP) models for mobile cameras can generate high-quality images that rival those of professional DSLR cameras. However, their computational demands often make them unsuitable for mobile settings. Additionally, modern mobile cameras employ non-Bayer color f...

Full description

Saved in:
Bibliographic Details
Main Authors: Minhyeok Cho, Haechang Lee, Hyunwoo Je, Kijeong Kim, Dongil Ryu, Albert No
Format: Article
Language:English
Published: IEEE 2023-01-01
Series:IEEE Access
Subjects:
Online Access:https://ieeexplore.ieee.org/document/10114931/
Tags: Add Tag
No Tags, Be the first to tag this record!
_version_ 1850066382868185088
author Minhyeok Cho
Haechang Lee
Hyunwoo Je
Kijeong Kim
Dongil Ryu
Albert No
author_facet Minhyeok Cho
Haechang Lee
Hyunwoo Je
Kijeong Kim
Dongil Ryu
Albert No
author_sort Minhyeok Cho
collection DOAJ
description Deep learning-based image signal processor (ISP) models for mobile cameras can generate high-quality images that rival those of professional DSLR cameras. However, their computational demands often make them unsuitable for mobile settings. Additionally, modern mobile cameras employ non-Bayer color filter arrays (CFA) such as Quad Bayer, Nona Bayer, and <inline-formula> <tex-math notation="LaTeX">$\text{Q}\times \text{Q}$ </tex-math></inline-formula> Bayer to enhance image quality, yet most existing deep learning-based ISP (or demosaicing) models focus primarily on standard Bayer CFAs. In this study, we present PyNET-<inline-formula> <tex-math notation="LaTeX">$\text{Q}\times \text{Q}$ </tex-math></inline-formula>, a lightweight demosaicing model specifically designed for <inline-formula> <tex-math notation="LaTeX">$\text{Q}\times \text{Q}$ </tex-math></inline-formula> Bayer CFA patterns, which is derived from the original PyNET. We also propose a knowledge distillation method called progressive distillation to train the reduced network more effectively. Consequently, PyNET-<inline-formula> <tex-math notation="LaTeX">$\text{Q}\times \text{Q}$ </tex-math></inline-formula> contains less than 2.5&#x0025; of the parameters of the original PyNET while preserving its performance. Experiments using <inline-formula> <tex-math notation="LaTeX">$\text{Q}\times \text{Q}$ </tex-math></inline-formula> images captured by a prototype <inline-formula> <tex-math notation="LaTeX">$\text{Q}\times \text{Q}$ </tex-math></inline-formula> camera sensor show that PyNET-<inline-formula> <tex-math notation="LaTeX">$\text{Q}\times \text{Q}$ </tex-math></inline-formula> outperforms existing conventional algorithms in terms of texture and edge reconstruction, despite its significantly reduced parameter count. Code and partial datasets can be found at <uri>https://github.com/Minhyeok01/PyNET-QxQ</uri>.
format Article
id doaj-art-9e2fe9950c4d475693b9042f412a593c
institution DOAJ
issn 2169-3536
language English
publishDate 2023-01-01
publisher IEEE
record_format Article
series IEEE Access
spelling doaj-art-9e2fe9950c4d475693b9042f412a593c2025-08-20T02:48:46ZengIEEEIEEE Access2169-35362023-01-0111448954491010.1109/ACCESS.2023.327266510114931PyNET-Q&#x00D7;Q: An Efficient PyNET Variant for Q&#x00D7;Q Bayer Pattern Demosaicing in CMOS Image SensorsMinhyeok Cho0https://orcid.org/0000-0002-2230-4473Haechang Lee1Hyunwoo Je2https://orcid.org/0000-0001-8014-0393Kijeong Kim3Dongil Ryu4Albert No5https://orcid.org/0000-0002-6346-4182Department of Electronic and Electrical Engineering, Hongik University, Seoul, South KoreaSK hynix, Icheon, South KoreaSK hynix, Icheon, South KoreaSK hynix, Icheon, South KoreaSK hynix, Icheon, South KoreaDepartment of Electronic and Electrical Engineering, Hongik University, Seoul, South KoreaDeep learning-based image signal processor (ISP) models for mobile cameras can generate high-quality images that rival those of professional DSLR cameras. However, their computational demands often make them unsuitable for mobile settings. Additionally, modern mobile cameras employ non-Bayer color filter arrays (CFA) such as Quad Bayer, Nona Bayer, and <inline-formula> <tex-math notation="LaTeX">$\text{Q}\times \text{Q}$ </tex-math></inline-formula> Bayer to enhance image quality, yet most existing deep learning-based ISP (or demosaicing) models focus primarily on standard Bayer CFAs. In this study, we present PyNET-<inline-formula> <tex-math notation="LaTeX">$\text{Q}\times \text{Q}$ </tex-math></inline-formula>, a lightweight demosaicing model specifically designed for <inline-formula> <tex-math notation="LaTeX">$\text{Q}\times \text{Q}$ </tex-math></inline-formula> Bayer CFA patterns, which is derived from the original PyNET. We also propose a knowledge distillation method called progressive distillation to train the reduced network more effectively. Consequently, PyNET-<inline-formula> <tex-math notation="LaTeX">$\text{Q}\times \text{Q}$ </tex-math></inline-formula> contains less than 2.5&#x0025; of the parameters of the original PyNET while preserving its performance. Experiments using <inline-formula> <tex-math notation="LaTeX">$\text{Q}\times \text{Q}$ </tex-math></inline-formula> images captured by a prototype <inline-formula> <tex-math notation="LaTeX">$\text{Q}\times \text{Q}$ </tex-math></inline-formula> camera sensor show that PyNET-<inline-formula> <tex-math notation="LaTeX">$\text{Q}\times \text{Q}$ </tex-math></inline-formula> outperforms existing conventional algorithms in terms of texture and edge reconstruction, despite its significantly reduced parameter count. Code and partial datasets can be found at <uri>https://github.com/Minhyeok01/PyNET-QxQ</uri>.https://ieeexplore.ieee.org/document/10114931/Bayer filtercolor filter array (CFA)demosaicingimage signal processor (ISP)knowledge distillationnon-Bayer CFA
spellingShingle Minhyeok Cho
Haechang Lee
Hyunwoo Je
Kijeong Kim
Dongil Ryu
Albert No
PyNET-Q&#x00D7;Q: An Efficient PyNET Variant for Q&#x00D7;Q Bayer Pattern Demosaicing in CMOS Image Sensors
IEEE Access
Bayer filter
color filter array (CFA)
demosaicing
image signal processor (ISP)
knowledge distillation
non-Bayer CFA
title PyNET-Q&#x00D7;Q: An Efficient PyNET Variant for Q&#x00D7;Q Bayer Pattern Demosaicing in CMOS Image Sensors
title_full PyNET-Q&#x00D7;Q: An Efficient PyNET Variant for Q&#x00D7;Q Bayer Pattern Demosaicing in CMOS Image Sensors
title_fullStr PyNET-Q&#x00D7;Q: An Efficient PyNET Variant for Q&#x00D7;Q Bayer Pattern Demosaicing in CMOS Image Sensors
title_full_unstemmed PyNET-Q&#x00D7;Q: An Efficient PyNET Variant for Q&#x00D7;Q Bayer Pattern Demosaicing in CMOS Image Sensors
title_short PyNET-Q&#x00D7;Q: An Efficient PyNET Variant for Q&#x00D7;Q Bayer Pattern Demosaicing in CMOS Image Sensors
title_sort pynet q x00d7 q an efficient pynet variant for q x00d7 q bayer pattern demosaicing in cmos image sensors
topic Bayer filter
color filter array (CFA)
demosaicing
image signal processor (ISP)
knowledge distillation
non-Bayer CFA
url https://ieeexplore.ieee.org/document/10114931/
work_keys_str_mv AT minhyeokcho pynetqx00d7qanefficientpynetvariantforqx00d7qbayerpatterndemosaicingincmosimagesensors
AT haechanglee pynetqx00d7qanefficientpynetvariantforqx00d7qbayerpatterndemosaicingincmosimagesensors
AT hyunwooje pynetqx00d7qanefficientpynetvariantforqx00d7qbayerpatterndemosaicingincmosimagesensors
AT kijeongkim pynetqx00d7qanefficientpynetvariantforqx00d7qbayerpatterndemosaicingincmosimagesensors
AT dongilryu pynetqx00d7qanefficientpynetvariantforqx00d7qbayerpatterndemosaicingincmosimagesensors
AT albertno pynetqx00d7qanefficientpynetvariantforqx00d7qbayerpatterndemosaicingincmosimagesensors