A High-Quality Denoising Dataset for Smartphone Cameras

By Abdelrahman Abdelhamed, Stephen Lin, Michael S. Brown. 2018 IEEE/CVF Conference on Computer Vision and Pattern Recognition, 2018

988 citations recorded by the source corpus.

The last decade has seen an astronomical shift from imaging with DSLR and point-and-shoot cameras to imaging with smartphone cameras. Due to the small aperture and sensor size, smartphone images have notably more noise than their DSLR counterparts. While denoising for smartphone images is an active research area, the research community currently lacks a denoising image dataset representative of real noisy images from smartphone cameras with high-quality ground truth. We address this issue in this paper with the following contributions. We propose a systematic procedure for estimating ground truth for noisy images that can be used to benchmark denoising performance for smartphone cameras. Using this procedure, we have captured a dataset - the Smartphone Image Denoising Dataset (SIDD) - of ~30,000 noisy images from 10 scenes under different lighting conditions using five representative smartphone cameras and generated their ground truth images. We used this dataset to benchmark a number of denoising algorithms. We show that CNN-based methods perform better when trained on our high-quality dataset than when trained using alternative strategies, such as low-ISO images used as a proxy for ground truth data.

Subfields: Computer Vision and Pattern Recognition, Media Technology

View Paper: https://www.semanticscholar.org/paper/ebf35073e122782f685a0d6c231622412f28a53b

How researchers see this paper

The researchers whose trusted networks rank it highest, strongest advocate first. Every researcher gets a different order.

In Computer Vision and Pattern Recognition

In Media Technology

Citations in the corpus

Deep Generalized Unfolding Networks for Image Restoration (2022, methodology): “Method SIDD [2] DND [52] PSNR↑ SSIM↑ PSNR↑ SSIM↑”
MambaIR: A Simple Baseline for Image Restoration with State-Space Model (2024, methodology): “For real image denoising, we train our model with 320 high-resolution images from SIDD [1] datasets, and use the SIDD test set and DND [58] dataset for testing.”
Simple Baselines for Image Restoration (2022, methodology): “31 dB of PSNR on SIDD[1]/GoPro[25] for image denoising/deblurring respectively.”
Variational Denoising Network: Toward Blind Noise Modeling and Removal (2019, methodology): “We compared the performance of VDN under different p values on the SIDD validation dataset in Table 6.”
Learning A Sparse Transformer Network for Effective Image Deraining (2023, methodology): “Table 1 reports results of different methods on the SIDD dataset [1].”
Generating Training Data for Denoising Real RGB Images via Camera Pipeline Simulation (2019, background): “[3] and [7] provide only RAW images, while [25] uses simple processing which may not be realistic.”
HINet: Half Instance Normalization Network for Image Restoration (2021, methodology): “28 dB in PSNR on SIDD [1] dataset, with only 7.”
Methods for image denoising using convolutional neural network: a review (2021, methodology): “They include: ImageNet large scale visual recognition challenge object detection (ILSVRC-DET) [122], Places2 [123], BerkeleySegmentation Data-set (BSD) [124], waterloo explorationdatabase [125], EMNIST [126], COCO dataset [127], MIT-Adobe five k [128], ImageNet [129], BSD68 [130], Set 14 [131], Reno”
Seeing Motion in the Dark (2019, methodology): “These include the RENOIR dataset [4], the Darmstadt Noise Dataset (DND) [37], the Smartphone Image Denoising Dataset (SIDD) [2], and the See in the Dark (SID) dataset [9].”
Restormer: Efficient Transformer for High-Resolution Image Restoration (2021, methodology): “Real image denoising on SIDD [1] and DND [60] datasets.”
MAXIM: Multi-Axis MLP for Image Processing (2022, methodology): “24 dB on SIDD [2]), 2) Deblurring (+0.”
Toward Convolutional Blind Denoising of Real Photographs (2018, background): “Fortunately, according to [43, 45, 1], nearly noisefree image can be obtained by averaging hundreds of noisy images from the same scene, and several datasets have been built in literatures.”
Spatial-Adaptive Network for Single Image Denoising (2020, positive): “ealize that real noise has a more complicated distribution. Thus, some recent works have focused on real noisy images. Several real noisy datasets have been established by capturing real noisy scenes [24, 3, 1], which promotes the research of real denoising methods. N3Net [25] demonstrated the signi”
Learning Enriched Features for Real Image Restoration and Enhancement (2020, methodology): “We train our network only on the training set of the SIDD [1] and directly evaluate it on the test images of both SIDD and DND [79] datasets.”
Images Speak in Images: A Generalist Painter for In-Context Visual Learning (2022, background): “Image restoration tasks are evaluated on several popular benchmarks, including SIDD [1] for image denoising, LoL [45] for low-light image enhancement, and the merged deraining dataset [51] for deraining.”
Neighbor2Neighbor: Self-Supervised Denoising from Single Noisy Images (2021, methodology): “For real-world experiments in raw-RGB space, we consider SIDD [2] dataset, a real-world denoising benchmark, which is collected using ﬁve smartphone cameras under 10 static scenes.”
Unprocessing Images for Learned Raw Denoising (2018, methodology): “As a result, recent work has demonstrated progress by collecting this real, paired data not just for evaluation, but for training models [1, 7].”
Spatial Attentive Single-Image Deraining With a High Quality Real Rain Dataset (2019, background): “Another approach to obtain a ground truth noise-free image is multi-frame fusion [46, 32, 1], which performs weighted averaging of a pre-aligned sequence of images taken from a static scene with a fixed camera setting.”