Raw Image Demoiréing Dataset

Citation Author(s):
Huanjing
Yue
Tianjin University
Yijia
Cheng
Tianjin University
Yan
Mao
Cong
Cao
Jingyu
Yang
Tianjin University
Submitted by:
Yijia Cheng
Last updated:
Wed, 08/10/2022 - 04:06
DOI:
10.21227/vkam-v640
Data Format:
License:
570 Views
Categories:
0
0 ratings - Please login to submit your rating.

Abstract 

We provide ground truth images and moiré images in raw domain and sRGB domain respectively, which are placed in four folders gt_RAW_npz, gt_RGB, moire_RAW_npz and moire_RGB. The ground truth raw image is actually pseudo ground truth. The users can regenerate them by utilizing other RGB to raw inversing algorithms. Our raw domain data is stored in npz format, including black level corrected data, black level value, white level value and white balance value.

Instructions: 

This dataset is constructed to learn the demoiriéng model via supervised learning. Please read our paper "Recaptured Screen Image Demoiréing in Raw Domain" published in IEEE Trans. on Multimedia for more details about the dataset construction and usage.

Funding Agency: 
National Natural Science Foundation of China
Grant Number: 
62072331

Comments

Our dataset is a benchmark for raw image demoiréing.

Submitted by Yijia Cheng on Wed, 08/10/2022 - 04:01

for low-light enchance research

Submitted by Jiang Feng on Thu, 04/11/2024 - 23:44