Single-Shot High Dynamic Range Imaging via Multiscale Convolutional Neural Network
We propose a single-shot high dynamic range (HDR) imaging algorithm with row-wise varying exposures in a single raw image based on a deep convolutional neural network (CNN). We first convert a raw Bayer input image into a radiance map by calibrating rows with different exposures, and then we design...
Main Authors: | , |
---|---|
Format: | Article |
Language: | English |
Published: |
IEEE
2021-01-01
|
Series: | IEEE Access |
Subjects: | |
Online Access: | https://ieeexplore.ieee.org/document/9427235/ |
id |
doaj-2fbc8218636848ceb4db6d0af38030b6 |
---|---|
record_format |
Article |
spelling |
doaj-2fbc8218636848ceb4db6d0af38030b62021-06-03T23:08:56ZengIEEEIEEE Access2169-35362021-01-019703697038110.1109/ACCESS.2021.30784579427235Single-Shot High Dynamic Range Imaging via Multiscale Convolutional Neural NetworkAn Gia Vien0https://orcid.org/0000-0003-0067-0285Chul Lee1https://orcid.org/0000-0001-9329-7365Department of Multimedia Engineering, Dongguk University, Seoul, South KoreaDepartment of Multimedia Engineering, Dongguk University, Seoul, South KoreaWe propose a single-shot high dynamic range (HDR) imaging algorithm with row-wise varying exposures in a single raw image based on a deep convolutional neural network (CNN). We first convert a raw Bayer input image into a radiance map by calibrating rows with different exposures, and then we design a new CNN model to restore missing information at the under- and over-exposed pixels and reconstruct color information from the raw radiance map. The proposed CNN model consists of three branch networks to obtain multiscale feature maps for an image. To effectively estimate the high-quality HDR images, we develop a robust loss function that considers the human visual system (HVS) model, color perception model, and multiscale contrast. Experimental results on both synthetic and captured real images demonstrate that the proposed algorithm can achieve synthesis results of significantly higher quality than conventional algorithms in terms of structure, color, and visual artifacts.https://ieeexplore.ieee.org/document/9427235/Spatially varying exposure (SVE) imagehigh dynamic range (HDR) imagingconvolutional neural network (CNN)human visual system (HVS) |
collection |
DOAJ |
language |
English |
format |
Article |
sources |
DOAJ |
author |
An Gia Vien Chul Lee |
spellingShingle |
An Gia Vien Chul Lee Single-Shot High Dynamic Range Imaging via Multiscale Convolutional Neural Network IEEE Access Spatially varying exposure (SVE) image high dynamic range (HDR) imaging convolutional neural network (CNN) human visual system (HVS) |
author_facet |
An Gia Vien Chul Lee |
author_sort |
An Gia Vien |
title |
Single-Shot High Dynamic Range Imaging via Multiscale Convolutional Neural Network |
title_short |
Single-Shot High Dynamic Range Imaging via Multiscale Convolutional Neural Network |
title_full |
Single-Shot High Dynamic Range Imaging via Multiscale Convolutional Neural Network |
title_fullStr |
Single-Shot High Dynamic Range Imaging via Multiscale Convolutional Neural Network |
title_full_unstemmed |
Single-Shot High Dynamic Range Imaging via Multiscale Convolutional Neural Network |
title_sort |
single-shot high dynamic range imaging via multiscale convolutional neural network |
publisher |
IEEE |
series |
IEEE Access |
issn |
2169-3536 |
publishDate |
2021-01-01 |
description |
We propose a single-shot high dynamic range (HDR) imaging algorithm with row-wise varying exposures in a single raw image based on a deep convolutional neural network (CNN). We first convert a raw Bayer input image into a radiance map by calibrating rows with different exposures, and then we design a new CNN model to restore missing information at the under- and over-exposed pixels and reconstruct color information from the raw radiance map. The proposed CNN model consists of three branch networks to obtain multiscale feature maps for an image. To effectively estimate the high-quality HDR images, we develop a robust loss function that considers the human visual system (HVS) model, color perception model, and multiscale contrast. Experimental results on both synthetic and captured real images demonstrate that the proposed algorithm can achieve synthesis results of significantly higher quality than conventional algorithms in terms of structure, color, and visual artifacts. |
topic |
Spatially varying exposure (SVE) image high dynamic range (HDR) imaging convolutional neural network (CNN) human visual system (HVS) |
url |
https://ieeexplore.ieee.org/document/9427235/ |
work_keys_str_mv |
AT angiavien singleshothighdynamicrangeimagingviamultiscaleconvolutionalneuralnetwork AT chullee singleshothighdynamicrangeimagingviamultiscaleconvolutionalneuralnetwork |
_version_ |
1721398544626089984 |