FFDNet: Toward a Fast and Flexible Solution for CNN-Based Image Denoising

By Kai Zhang, Wangmeng Zuo, Lei Zhang. 2017

2,636 citations recorded by the source corpus.

Due to the fast inference and good performance, discriminative learning methods have been widely studied in image denoising. However, these methods mostly learn a specific model for each noise level, and require multiple models for denoising images with different noise levels. They also lack flexibility to deal with spatially variant noise, limiting their applications in practical denoising. To address these issues, we present a fast and flexible denoising convolutional neural network, namely FFDNet, with a tunable noise level map as the input. The proposed FFDNet works on downsampled sub-images, achieving a good trade-off between inference speed and denoising performance. In contrast to the existing discriminative denoisers, FFDNet enjoys several desirable properties, including: 1) the ability to handle a wide range of noise levels (i.e., [0, 75]) effectively with a single network; 2) the ability to remove spatially variant noise by specifying a non-uniform noise level map; and 3) faster speed than benchmark BM3D even on CPU without sacrificing denoising performance. Extensive experiments on synthetic and real noisy images are conducted to evaluate FFDNet in comparison with state-of-the-art denoisers. The results show that FFDNet is effective and efficient, making it highly attractive for practical denoising applications.

Subfields: Computer Vision and Pattern Recognition, Control and Systems Engineering

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 Control and Systems Engineering

Citations in the corpus

Designing a Practical Degradation Model for Deep Blind Image Super-Resolution (2021, background): “types and the real RealSRSet which consists of 20 real images either downloaded from the internet or directly chosen from existing testing datasets [17, 32, 33, 52].”
Memory-Efficient Hierarchical Neural Architecture Search for Image Denoising (2019, background): “Lately, FFDNet [43] attempts to address spatially varying noise by appending noise level maps to the input of DnCNN.”
Pre-Trained Image Processing Transformer (2020, neutral): “…of the proposed method, Urban100( × 4): img 004 HR VDSR[41] EDSR[51] RDN[94] OISR[35] SAN[17] RNAN [93] IGNN BSD68: 163085 GT Noisy ( σ =50) CBM3D [16] TNRD [14] RDN [94] DnCNN [87] MemNet [65] IRCNN [88] FFDNet [89] IPT (ours) we compare our results with various state-of-the-art models.”
Generating Training Data for Denoising Real RGB Images via Camera Pipeline Simulation (2019, methodology): “N3Net [26] formulates a differentiable version of nearest neighbor search to further improve DnCNN. FFDNet [33] attempts to address spatially varying noise by appending noise level maps to the input of DnCNN.”
All-In-One Image Restoration for Unknown Corruption (2022, methodology): “Method CBM3D [5] DnCNN [53] IRCNN [54] FFDNet [55] BRDNet [40] DL [8] Ours”
Neural Nearest Neighbors Networks (2018, methodology): “We compare our full N 3 Net against state-of-the-art local denoising methods, i. e. the DnCNN baseline [51], the very deep and wide ( 30 layers, 128 feature channels) RED30 model [31], and the recent FFDNet [53].”
Attention-guided CNN for image denoising (2020, methodology): “For example, a fast and ﬂexible denoising convolutional neural network (FFDNet) [59] used the noisy image patches and the noise mapping patches as input of the network to efﬁciently tackle blind denoising. e Further, separating the convolutions can improve the denoising efﬁciency [7]. r From these d”
SwinIR: Image Restoration Using Swin Transformer (2021, methodology): “Noisy BM3D [14] DnCNN [90] FFDNet [92] DRUNet [88] SwinIR (ours)”
Deep Unfolding Network for Image Super-Resolution (2020, methodology): “Inspired by [66], we propose a deep CNN denoiser that takes the noise level as input”
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.”
IR-NAS: Neural Architecture Search for Image Restoration (2019, background): “Lately, FFDNet [6] attempts to address spatially varying noise by appending noise level maps to the input of DnCNN. NLRN [7] incorporates non-local operations into a recurrent neural network (RNN) for image restoration.”
R2RNet: Low-light Image Enhancement via Real-low to Real-normal Network (2021, methodology): “Image denoising methods based on supervised learning, such as DnCNN-B [30], FFDNet [31], and CBDNet [32] utilized the Gaussian mixture model to perform denoising.”
Noise2Void - Learning Denoising From Single Noisy Images (2018, neutral): “In recent years, convolutional neural networks (CNNs) have been trained in various ways to predict pixel values from surrounding image patches, i. tive ﬁeld of that pixel [24, 11, 26, 6, 23, 25, 18, 14].”
Dual Residual Networks Leveraging the Potential of Paired Operations for Image Restoration (2019, background): “Gaussian noise removal Application of neural networks to noise removal has a long history [1,18,43,53,54].”
Blind Face Restoration via Deep Multi-scale Component Dictionaries (2020, background): “, image super-resolution [9,19,24,44, 46], denoising [13, 38, 42, 43], deblurring [22, 29, 41], and compression artifact removal [8, 10, 12].”
Deep Learning on Image Denoising: An overview (2019, positive): “An FFDNet [260] used noise level and noise as the input of CNN to train a denoiser for unknown noisy images.”
User-Centric Cell-Free Massive MIMO Networks: A Survey of Opportunities, Challenges and Solutions (2021, methodology): “We also plot the results for a manually modiﬁed PLE of 2 .”