Home

Awesome

MC-Blur: A Comprehensive Benchmark for Image Deblurring

Our propsoed MC-Blur Benchmark

We construct a large-scale multi-cause (MC-Blur) dataset for image deblurring. It consists of four blur types: uniform blurs, motion blurs by averaging continuous frames, heavy defocus blurs, and real-world blurs. We collect these images from more than 1000 diverse scenes such as buildings, city scenes, vehicles, natural landscapes, people, animals, and sculptures. MC-Blur Benchmark consits of four different subsets, i.e., Real high-fps based Motion-blurred subset (RHM), large-kernel UHD Motion-blurred subset (UHDM), large-scale heavy defocus blurred subset (LSD), and Real Mixed Blurry Qualitative subset (RMBQ).

Downloads

The images of the dataset can be downloaded from the links below.

Google Drive

<!-- - [RHM-250](https://drive.google.com/file/d/1hoCFNeP1GOszaJfLABBw35hCQQKWPMhV/view?usp=sharing) (6.2G for test) --> <!-- - [RHM-500](https://drive.google.com/file/d/13payJmIY6mssFXMSuOwf1aosA3n6bXb6/view?usp=sharing) (7.1G for test) --> <!-- - [RHM-1000](https://drive.google.com/file/d/1HJqV6Ogve-G6YGYJDH7MD5nKAQOTnVMG/view?usp=sharing) (9.3G for test) -->

Baidu Cloud (How to unzip?)

Download MC-Blur benchmark from the script, run

python download_data.py

Note: The above script will download all subsets of the MC-Blur. You can use "--data" to select. For example:

python download_data.py --data "UHDM_train_test"

Some visual examples of MC-Blur Dataset

Visual examples for each subset of our MC-Blur Dataset.

Some code steps in synthesizing dataset

See detail in README.

Benchmarking Study

Methods

DatePublicationTitleAbbreviationCodePlatform
2017CVPRDeep multi-scale convolutional neural network for dynamic scene deblurring paperDeepDeblurCodePytorch
2018CVPRDeblurgan: Blind motion deblurring using conditional adversarial networks paperDeblurGANCodePytorch
2018CVPRScale-recurrent network for deep image deblurring paperSRNCodeTensorflow
2019ICCVDeblurGAN-v2: Deblurring (Orders-of-Magnitude) Faster and Better paperDeblurGAN-v2CodePytorch
2019CVPRDeep Stacked Hierarchical Multi-Patch Network for Image Deblurring paperDMPHNCodePytorch
2020CVPRDeblurring by Realistic Blurring paperDBGANCodePytorch
2021CVPRMulti-Stage Progressive Image Restoration paperMPRNetCodePytorch
2022CVPRRestormer: Efficient Transformer for High-Resolution Image Restoration paperRestormerCodePytorch
2021ICCVRethinking Coarse-To-Fine Approach in Single Image Deblurring paperMIMO-UNetCodePytorch

Metrics

AbbreviationFull-/Non-ReferencePlatformCode
PSNR (Peak Signal-to-Noise Ratio)Full-Reference
SSIM (Structural Similarity Index Measurement)Full-ReferenceMATLABCode
NIQE (Naturalness Image Quality Evaluator)Non-ReferenceMATLABCode
SSEQ (No-reference Image Quality Assessment Based on Spatial and Spectral Entropies)Non-ReferenceMATLABCode

Results for 250-fps images from RHM Set

MethodPSNRSSIMParameter
DeepDeblur30.380.876611.72 M
DeblurGAN24.890.63646.07 M
SRN30.570.87996.88 M
DeblurGAN-v226.990.80617.84 M
DMPHN30.420.876821.69 M
DBGAN27.890.819111.59 M
MPRNet31.520.923920.13 M
Restormer30.410.910626.10 M
MIMO-UNet32.020.92856.81 M

Results for 500-fps images from RHM Set

MethodPSNRSSIMParameter
DeepDeblur31.080.897411.72 M
DeblurGAN24.660.67486.07 M
SRN31.540.90516.88 MB
DeblurGAN-v227.670.83207.84 M
DMPHN31.430.901821.69 M
DBGAN28.360.838811.59 M
MPRNet32.080.930020.13 M
Restormer30.980.916026.10 M
MIMO-UNet32.890.93986.81 M

Results for 1000-fps images from RHM Set

MethodPSNRSSIMParameter
DeepDeblur32.410.896611.72 M
DeblurGAN25.200.65356.07 M
SRN32.690.0.90166.88 M
DeblurGAN-v229.810.84617.84 M
DMPHN32.410.909621.69 M
DBGAN29.660.831811.59 M
MPRNet33.360.933220.13 M
Restormer32.770.926426.10 M
MIMO-UNet33.750.93606.81 M

Results on UHDM Set

MethodPSNRSSIMParameter
DeepDeblur22.230.632211.72 M
DeblurGAN20.390.55686.07 M
SRN22.280.63466.88 M
DeblurGAN-v221.030.58397.84 M
DMPHN22.200.637821.69 M
DBGAN21.520.602511.59 M
MPRNet23.700.747220.13 M
Restormer22.390.735626.10 M
MIMO-UNet22.970.73176.81 M

Results on LSD Set

MethodPSNRSSIMParameter
DeepDeblur20.730.721811.72 M
DeblurGAN20.040.63356.07 M
SRN21.660.76646.88 M
DeblurGAN-v221.130.69647.84 M
DMPHN21.230.751921.69 M
DBGAN21.560.753611.59 M
MPRNet21.320.789720.13 M
Restormer22.350.807226.10 M
MIMO-UNet22.560.79856.81 M

Citation

If you think this work is useful for your research, please cite the following paper.

@inproceedings{zhang2023benchmarking,
  title={MC-Blur: A Comprehensive Benchmark for Image Deblurring},
  author={Zhang, Kaihao and Wang, Tao and Luo, Wenhan and Chen, Boheng and Ren, Wenqi and Stenger, Bjorn and Liu, Wei and Li, Hongdong and Yang Ming-Hsuan},
  booktitle={IEEE Transactions on Circuits and Systems for Video Technology},
  year={2023}
}

License

The MC-Blur dataset is released under CC BY-NC-ND license.