Toward Convolutional Blind Denoising of Real Photographs

By Shi Guo, Zifei Yan, Kai Zhang, Wangmeng Zuo, Lei Zhang. 2018

1,113 citations recorded by the source corpus.

While deep convolutional neural networks (CNNs) have achieved impressive success in image denoising with additive white Gaussian noise (AWGN), their performance remains limited on real-world noisy photographs. The main reason is that their learned models are easy to overfit on the simplified AWGN model which deviates severely from the complicated real-world noise model. In order to improve the generalization ability of deep CNN denoisers, we suggest training a convolutional blind denoising network (CBDNet) with more realistic noise model and real-world noisy-clean image pairs. On the one hand, both signal-dependent noise and in-camera signal processing pipeline is considered to synthesize realistic noisy images. On the other hand, real-world noisy photographs and their nearly noise-free counterparts are also included to train our CBDNet. To further provide an interactive strategy to rectify denoising result conveniently, a noise estimation subnetwork with asymmetric learning to suppress under-estimation of noise level is embedded into CBDNet. Extensive experimental results on three datasets of real-world noisy pho- tographs clearly demonstrate the superior performance of CBDNet over state-of-the-arts in terms of quantitative met- rics and visual quality. The code has been made available at https://github.com/GuoShi28/CBDNet.

Subfields: Computer Vision and Pattern Recognition, Media Technology

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

Memory-Efficient Hierarchical Neural Architecture Search for Image Denoising (2019, methodology): “CBDNet [10] uses a simulated camera pipeline to supplement real training data.”
Guided Collaborative Training for Pixel-wise Semi-Supervised Learning (2020, background): “Real image denoising [3,15,50] is a task that devotes to removing the real noise, rather synthetic noise, from an input natural image (Fig.”
MADNet: A Fast and Lightweight Network for Single-Image Super Resolution (2020, background): “Given the perception that MAE or MSE tends to lead a smooth result, we additionally introduce a total variation (TV) regularizer [10], [29] to constrain the smoothness of where ∇ h ( · ) and ∇ v ( · ) denote the gradient operator among the horizontal and vertical direction, respectively.”
Variational Denoising Network: Toward Blind Noise Modeling and Removal (2019, methodology): “Comparison methods: Several state-of-the-art denoising methods are adopted for performance comparison, including CBM3D [11], WNNM [16], NCSR [14], MLP [10], DnCNN-B [40], FFDNet [41], UDNet [24] and CBDNet [17].”
Neighbor2Neighbor: Self-Supervised Denoising from Single Noisy Images (2021, background): “After that, numerous denoising networks are proposed to further improve the performance [20, 28, 35, 16, 24, 12, 10].”
IR-NAS: Neural Architecture Search for Image Restoration (2019, methodology): “CBDNet [9] uses simulated camera pipeline to supplement real training data.”
Spatial-Adaptive Network for Single Image Denoising (2020, positive): “ith JPEG compression. Furthermore, as comparisons, we choose the state-of-the-art methods whose validity has previously been demonstrated on real noisy images, including CBM3D [8], DnCNN [35], CBDNet [12], PD [39], and RIDNet [4]. DND The quantitative results are listed in Table 4, which are obtaine”
Learning Enriched Features for Fast Image Restoration and Enhancement (2022, methodology): “Recently, deep learning models [6], [9], [32], [33], [34], [35], [36], [37], [38], [39], [40], [41], [42] make significant advances in image denoising, yielding favorable results than those of the hand-crafted methods.”
High-Quality Self-Supervised Deep Image Denoising (2019, background): “Real corrupted images rarely follow theoretical models exactly [10, 18, 25], and an important avenue for future work will be to learn as much of the noise model from the data as possible.”
Blind Face Restoration via Deep Multi-scale Component Dictionaries (2020, methodology): “. 2.1 Single Image Restoration Along with the benets brought by deep CNNs, single image restoration has achieved great success in many tasks, e.g., image super-resolution [9,19,24,44, 46], denoising [13,38,42,43], deblurring [22,29,41], and compression artifact removal [8,10,12]. Due to the specic f”
CycleISP: Real Image Restoration via Improved Data Synthesis (2020, methodology): “Several recent methods use deep CNNs [4, 7, 25, 28, 45, 64, 65, 2] and demonstrate promising denoising performance.”
GAN Prior Embedded Network for Blind Face Restoration in the Wild (2021, background): “Recently, deep neural networks (DNNs) have shown superior results in a variety of computer vision tasks [24, 48, 13, 25, 30], and many DNN based face restoration methods [49, 29, 16] have also been developed and they have demonstrated much better performance than traditional ones.”
Towards Flexible Blind JPEG Artifacts Removal (2021, methodology): “Cascaded QF prediction and non-blind model: It is also possible to design a QF predictor cascaded by a nonblind method like CBD-Net [18].”
Neural Nearest Neighbors Networks (2018, methodology): “Moreover, the PSNR values, when evaluated on developed sRGB images, surpass those of the currently top performing methods in sRGB denoising, TWSC [46] and CBDNet [15].”
Learning Enriched Features for Real Image Restoration and Enhancement (2020, methodology): “Noisy CBDNet [39] RIDNet [5] VDN [118] MIRNet (Ours) Reference”
FastDVDnet: Towards Real-Time Deep Video Denoising Without Flow Estimation (2019, methodology): “In other cases, such as JPEG denoising, the noise map can be estimated by means of an additional CNN [20].”
Unprocessing Images for Learned Raw Denoising (2018, methodology): “Our model is the fastest technique by a significant margin: 9× faster than N3Net [32] and 18× faster than CBDnet [18], the next two best performing techniques after our own.”
Learning to Restore Low-Light Images via Decomposition-and-Enhancement (2020, methodology): “images, by synthesizing noise in the raw data domain [2], constructing real image dataset [1], developing joint training strategy of both synthetic and real images [19], or unsupervised learning [28].”