Flow-based Kernel Prior with Application to Blind Super-Resolution
By Jingyun Liang, Kai Zhang, Shuhang Gu, Luc Van Gool, Radu Timofte. 2021
155 citations recorded by the source corpus.
Kernel estimation is generally one of the key problems for blind image super-resolution (SR). Recently, Double-DIP proposes to model the kernel via a network architecture prior, while KernelGAN employs the deep linear network and several regularization losses to constrain the kernel space. However, they fail to fully exploit the general SR kernel assumption that anisotropic Gaussian kernels are sufficient for image SR. To address this issue, this paper proposes a normalizing flow-based kernel prior (FKP) for kernel modeling. By learning an invertible mapping between the anisotropic Gaussian kernel distribution and a tractable latent distribution, FKP can be easily used to replace the kernel modeling modules of Double-DIP and KernelGAN. Specifically, FKP optimizes the kernel in the latent space rather than the network parameter space, which allows it to generate reasonable kernel initialization, traverse the learned kernel manifold and improve the optimization stability. Extensive experiments on synthetic and real-world images demonstrate that the proposed FKP can significantly improve the kernel estimation accuracy with less parameters, runtime and memory usage, leading to state-of-the-art blind SR results.
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
- Luc Van Gool
- Radu Timofte
- Chen Change Loy
- Ming-Hsuan Yang
- Kai Zhang
- Lei Zhang
- Yu Qiao
- Martin Danelljan
In Media Technology
- Radu Timofte
- Luc Van Gool
- Chen Change Loy
- Kai Zhang
- Jingyun Liang
- Chao Dong
- Zongsheng Yue
- Jianyi Wang
Citations in the corpus
- Designing a Practical Degradation Model for Deep Blind Image Super-Resolution (2021, methodology): “As mentioned in the introduction, existing DNNs-based SISR methods are generally based on bicubic downsam-pling [23, 44] and traditional degradations [26, 37, 54, 59, 60], or some simple variants [11, 41, 53, 56, 58].”
- Hierarchical Conditional Flow: A Unified Framework for Image Super-Resolution and Image Rescaling (2021, methodology): “Flow-based models [7, 8, 23, 37, 13, 22, 2, 16, 12, 36, 28] aim to learn a bijective mapping between the target space and the latent space.”
- ResShift: Efficient Diffusion Model for Image Super-resolution by Residual Shifting (2023, background): “Various SR technologies were explored from different perspectives, including network architecture [31, 32, 33, 34], image prior [35, 36, 37, 38], deep unfolding [39, 40, 41], degradation model [17, 18, 42, 43].”
- SwinIR: Image Restoration Using Swin Transformer (2021, unknown): “Since pioneering work SRCNN [18] (for image SR), DnCNN [90] (for image denoising) and ARCNN [17] (for JPEG compression artifact reduction), a flurry of CNN-based models have been proposed to improve model representation ability by using larger and deeper neural network architecture designs, such as ”