Blind Image Deblurring Using Dark Channel Prior

By Jin-shan Pan, Deqing Sun, Hanspeter Pfister, Ming-Hsuan Yang. Computer Vision and Pattern Recognition, 2016

785 citations recorded by the source corpus.

We present a simple and effective blind image deblurring method based on the dark channel prior. Our work is inspired by the interesting observation that the dark channel of blurred images is less sparse. While most image patches in the clean image contain some dark pixels, these pixels are not dark when averaged with neighboring highintensity pixels during the blur process. This change in the sparsity of the dark channel is an inherent property of the blur process, which we both prove mathematically and validate using training data. Therefore, enforcing the sparsity of the dark channel helps blind deblurring on various scenarios, including natural, face, text, and low-illumination images. However, sparsity of the dark channel introduces a non-convex non-linear optimization problem. We introduce a linear approximation of the min operator to compute the dark channel. Our look-up-table-based method converges fast in practice and can be directly extended to non-uniform deblurring. Extensive experiments show that our method achieves state-of-the-art results on deblurring natural images and compares favorably methods that are well-engineered for specific scenarios.

Subfields: Computer Vision and Pattern Recognition, Media Technology

View Paper: https://www.semanticscholar.org/paper/6584cdd5ac482759a7d5f75f45b690e92fbe6185

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

Dynamic Scene Deblurring With Parameter Selective Sharing and Nested Skip Connections (2019, background): “Many natural image priors [1, 27, 2, 35, 36, 21, 22] were proposed to regularize the solution space and advance the deblurring research.”
Dynamic Scene Deblurring Using Spatially Variant Recurrent Neural Networks (2018, methodology): “(a) blurry image (b) Whyte [36] (c) Xu [41] (d) Sun [33] (e) Pan [24] psnr/ssim 27.”
DeblurGAN-v2: Deblurring (Orders-of-Magnitude) Faster and Better (2019, methodology): “[36], DeepDeblur [33], SRN [45], DeblurGAN [21]; and the three DeblurGAN-v2 variants (Inception-ResNet-v2, Mo-bileNet, MobileNet-DSC).”
Deep Unfolding Network for Image Super-Resolution (2020, neutral): “The two kernels are chosen from [41].”
Learning to Super-Resolve Blurry Face and Text Images (2017, methodology): “This is in sharp contrast to empirical priors [9, 20, 23, 32, 33, 45] that are developed using the statistics of natural images.”
Image Deblurring via Extreme Channels Prior (2017, methodology): “We combine the proposed ECP based regularization term (9) into the recent single image deblurring framework in [17], then our proposed objective function for image deblurring becomes”
Scale-Recurrent Network for Deep Image Deblurring (2018, background): “uniform/non-uniform/depth-aware), and utilized different natural image priors [1, 3, 6, 39, 14, 40, 26, 27] to regularize the solution space.”
Meta-Transfer Learning for Zero-Shot Super-Resolution (2020, methodology): “With a given LR image, we downsample it with corresponding downsampling kernel (kernel estimation algorithms [24, 29] can be adopted for blind scenario) to generate Ison and perform a few gradient updates with respect to the model parameter using a single pair of “LR son” and a given image.”
Human-Aware Motion Deblurring (2019, background): “Blind-Image Deblurring: For the uniform blur problem, conventional methods typically resort to natural image priors to estimate latent images [8, 21, 36, 23, 32, 28, 51, 5, 14, 41, 50, 53, 35].”
Learning Filter Basis for Convolutional Neural Network Compression (2019, background): “Recently, deep convolutional neural network (CNN) based approaches have been setting new state-of-the-art results not only for high-level computer vision tasks such as image classification [25, 44, 15, 18], segmentation [11, 35], and object detection [11, 10, 43, 42], but also for low-level tasks su”
Spatio-Temporal Filter Adaptive Network for Video Deblurring (2019, background): “Early researchers assume a uniform blur kernel and design some natural image priors, such as L0-regularized prior [43], dark channel prior [28], to compensate for the ill-posed blur removal process.”
Unsupervised Domain-Specific Deblurring via Disentangled Representations (2019, methodology): “Conventional methods [28, 29, 27, 41, 14] still tend to under-deblur or over-deblur images, especially on local regions such as eyes and mouths.”
Missing Data Reconstruction in Remote Sensing Image With a Unified Spatial–Temporal–Spectral Deep Convolutional Neural Network (2018, background): “3) Dilated Convolution: In image inverse problems such as image inpainting [38], denoising [39], [40], and deblurring [41], contextual information can effectively promote the Fig.”
Real-World Blur Dataset for Learning and Benchmarking Deblurring Algorithms (2020, methodology): “Thus, image deblurring, a problem to restore a sharp image from a blurred one, has been extensively studied for decades [12, 38, 7, 45, 46, 33, 40, 6].”
From Motion Blur to Motion Flow: A Deep Learning Solution for Removing Heterogeneous Motion Blur (2016, background): “Numerous image priors or regularizers have been studied to overcome the ill-posed nature of the problem, such as the total variational regularizer [5, 29], Gaussian scale mixture priors [9] and 1/ 2-norms [19], 0-norms [44, 27], and dark channel [28] based regularizers.”