122 Hits in 3.6 sec

Wasserstein Patch Prior for Image Superresolution [article]

Johannes Hertrich, Antoine Houdard, Claudia Redenbach
2021 arXiv   pre-print
In this paper, we introduce a Wasserstein patch prior for superresolution of two- and three-dimensional images.  ...  Then, the proposed regularizer penalizes the W_2-distance of the patch distribution of the reconstruction to the patch distribution of some reference image at different scales.  ...  Inspired by [20], we propose in this paper to use the Wasserstein-2 distance from the patch distribution of the reconstruction to the patch distribution of our reference image as a patch prior for superresolution  ... 
arXiv:2109.12880v2 fatcat:3ccrd36pqzabppkveahukkb3te

WPPNets and WPPFlows: The Power of Wasserstein Patch Priors for Superresolution [article]

Fabian Altekrüger, Johannes Hertrich
2022 arXiv   pre-print
Recently, Wasserstein patch priors (WPP), which are based on the comparison of the patch distributions of the unknown image and a reference image, were successfully used as data-driven regularizers in  ...  Numerical examples demonstrate the very good performance of WPPNets for superresolution in various image classes even if the forward operator is known only approximately.  ...  Based on the empirical patch distributions of an image x and a reference image x, we define the Wasserstein Patch Prior (WPP) as the squared Wasserstein-2 distance of the corresponding empirical patch  ... 
arXiv:2201.08157v2 fatcat:hsmro6chfzcgjpahf352slomiu

WPPNets: Unsupervised CNN Training with Wasserstein Patch Priors for Image Superresolution [article]

Fabian Altekrüger, Johannes Hertrich
Then, we propose a loss function based on the Wasserstein patch prior which measures the Wasserstein-2 distance between the patch distributions of the predictions and the reference image.  ...  We introduce WPPNets, which are CNNs trained by a new unsupervised loss function for image superresolution of materials microstructures.  ...  Wasserstein patch prior for image superresolution. arXiv preprint arXiv:2109.12880, 2021. [19] J. Hertrich, S. Neumayer, and G. Steidl.  ... 
doi:10.48550/arxiv.2201.08157 fatcat:daobkrxnb5anfigefdlopvnnxu

PatchNR: Learning from Small Data by Patch Normalizing Flow Regularization [article]

Fabian Altekrüger, Alexander Denker, Paul Hagemann, Johannes Hertrich, Peter Maass, Gabriele Steidl
2022 arXiv   pre-print
Numerical examples for low-dose CT, limited-angle CT and superresolution of material images demonstrate that our method provides high quality results among unsupervised methods, but requires only few data  ...  Our regularizer, called patchNR, involves a normalizing flow learned on patches of very few images.  ...  The idea of the Wasserstein Patch Prior (WPP) [3, 22] 10 is to use the Wasserstein-2 distance between the patch distribution of the reconstruction and the patch distribution of a given reference image  ... 
arXiv:2205.12021v1 fatcat:wv22j6q3mjh5bg76lmw43jczli

Generative Adversarial Networks for Image Super-Resolution: A Survey [article]

Chunwei Tian, Xuanyu Zhang, Jerry Chun-Wen Lin, Wangmeng Zuo, Yanning Zhang
2022 arXiv   pre-print
Second, we present popular architectures for GANs in big and small samples for image applications.  ...  Single image super-resolution (SISR) has played an important role in the field of image processing.  ...  Besides, Wasserstein distance is used to enhance the stability of training a remote sensing superresolution model [84] .  ... 
arXiv:2204.13620v1 fatcat:hlwdqith65cxrbqrnbphjz6u4u

Wasserstein Loss for Image Synthesis and Restoration

Guillaume Tartavel, Gabriel Peyré, Yann Gousseau
2016 SIAM Journal of Imaging Sciences  
This paper presents a novel variational approach to impose statistical constraints to the output of both image generation (to perform typically texture synthesis) and image restoration (for instance to  ...  For applications to texture synthesis, the input distributions are the empirical distributions computed from an exemplar image.  ...  Wasserstein Loss for Image Restoration.  ... 
doi:10.1137/16m1067494 fatcat:vrbtvkvbxfdrplr42sk632ob5i

Breast Cancer Histopathology Image Super-Resolution Using Wide-Attention GAN with Improved Wasserstein Gradient Penalty and Perceptual Loss

Faezehsadat Shahidi
2021 IEEE Access  
In the realm of image processing, enhancing the quality of the images is known as a superresolution problem (SR).  ...  As it is of the utmost importance to keep the size and the shape of the images, while enlarging the medical images, we propose a novel superresolution model with a generative adversarial network to generate  ...  Mohsen Golshiripour for his support with financial aid. We wish to acknowledge Dr. Sebelan Danishvar and Mr. Behnam Kiani Kalejahi for their support and assistance with resources.  ... 
doi:10.1109/access.2021.3057497 fatcat:wltgx3gngrei5jiglnzy2rcvki

Super-resolution MRI through Deep Learning [article]

Qing Lyu, Chenyu You, Hongming Shan, Ge Wang
2018 arXiv   pre-print
Magnetic resonance imaging (MRI) is extensively used for diagnosis and image-guided therapeutics.  ...  Compared with polynomial interpolation or sparse-coding algorithms, deep learning extracts prior knowledge from big data and produces superior MRI images from a low-resolution counterpart.  ...  Training Process During the training process, the LR and HR images were divided into 64 × 64 patches. The training process was conducted in the supervised learning framework.  ... 
arXiv:1810.06776v1 fatcat:46mnpqxffnd23ky3wzurkzm6si

A Novel Medical Image Denoising Method Based on Conditional Generative Adversarial Network

Yuqin Li, Ke Zhang, Weili Shi, Yu Miao, Zhengang Jiang, Cristiana Corsi
2021 Computational and Mathematical Methods in Medicine  
Most existing methods for medical image denoising adapted to certain types of noise have difficulties in handling spatially varying noise; meanwhile, image detail losses and structure changes occurred  ...  Considering image context perception and structure preserving, this paper firstly introduces a medical image denoising method based on conditional generative adversarial network (CGAN) for various unknown  ...  a batch of raw image patches I raw and the image to be processed patches I grad aug 9.  ... 
doi:10.1155/2021/9974017 pmid:34621329 pmcid:PMC8492295 fatcat:eqxotnbmx5fdpirxof35axptsy

Unpaired Image Super-Resolution with Optimal Transport Maps [article]

Milena Gazdieva, Litu Rout, Alexander Korotin, Alexander Filippov, Evgeny Burnaev
2022 arXiv   pre-print
Inspired by these findings, we propose an algorithm for unpaired SR which learns an unbiased OT map for the perceptual transport cost.  ...  Second, we empirically show that the learned map is biased, i.e., it may not actually transform the distribution of low-resolution images to high-resolution images.  ...  FID is computed on 128 × 128 patches of LR test images upsampled by the method in view w.r.t. random patches of test HR images. We use 50000 patches to compute FID.  ... 
arXiv:2202.01116v1 fatcat:fzr5b43kyze67me2kqr2b2zraa

A Latent Encoder Coupled Generative Adversarial Network (LE-GAN) for Efficient Hyperspectral Image Super-resolution [article]

Yue Shi, Liangxiu Han, Lianghao Han, Sheng Chang, Tongle Hu, Darren Dancey
2021 arXiv   pre-print
The generative adversarial network (GAN) has proven to be an effective deep learning framework for image super-resolution.  ...  AVIRIS and UHD-185) for various upscaling factors and added noise levels, and compared with the state-of-the-art super-resolution models (i.e. HyCoNet, LTTR, BAGAN, SR- GAN, WGAN).  ...  For further information, please visit  ... 
arXiv:2111.08685v1 fatcat:xfabty4425f2ljiwdjpwyonxd4

Online Video Super-Resolution with Convolutional Kernel Bypass Graft [article]

Jun Xiao, Xinyang Jiang, Ningxin Zheng, Huan Yang, Yifan Yang, Yuqing Yang, Dongsheng Li, Kin-Man Lam
2022 arXiv   pre-print
pretrained image SR models.  ...  Then, our proposed CKBG method enhances this lightweight base model by bypassing the original network with "kernel grafts", which are extra convolutional kernels containing the prior knowledge of external  ...  The extracted kernels contain a lot of prior information for generating HR images.  ... 
arXiv:2208.02470v1 fatcat:sk3jsz4g6vbmngtsw3lk33sdl4

Super-resolution emulator of cosmological simulations using deep physical models [article]

Doogesh Kodi Ramanah, Tom Charnock, Francisco Villaescusa-Navarro, Benjamin D. Wandelt
2020 arXiv   pre-print
We present an extension of our recently developed Wasserstein optimized model to emulate accurate high-resolution features from computationally cheaper low-resolution cosmological simulations.  ...  We exploit the information content of the high-resolution initial conditions as a well constructed prior distribution from which the network emulates the small-scale structures.  ...  ACKNOWLEDGEMENTS We thank the reviewer for their constructive feedback which helped to improve the quality of the manuscript. We express our appreciation to Guilhem Lavaux for his valuable insights.  ... 
arXiv:2001.05519v2 fatcat:xje7pwwbp5hvvl3uua2sg7cv7y

HypervolGAN: An efficient approach for GAN with multi-objective training function [article]

Jingwen Su, Hujun Yin
2020 arXiv   pre-print
For instance, in image enhancement or restoration, there are often several criteria to consider such as signal-noise ratio, smoothness, structures and details.  ...  We tested our proposed method on solving single image super-resolution problem.  ...  Domain-based SISR algorithms use specific class of image priors [35, 36] , while generic SISR algorithms use general image priors like edges [17] , image statistics (e.g. heavy-tailed gradient distribution  ... 
arXiv:2006.15228v1 fatcat:fddye725n5fmpat4xq6kvidxmy

NTIRE 2020 Challenge on Image and Video Deblurring

Seungjun Nah, Sanghyun Son, Radu Timofte, Kyoung Mu Lee, Yu Tseng, Yu-Syuan Xu, Cheng-Ming Chiang, Yi-Min Tsai, Stephan Brehm, Sebastian Scherer, Dejia Xu, Yihao Chu (+36 others)
2020 2020 IEEE/CVF Conference on Computer Vision and Pattern Recognition Workshops (CVPRW)  
Track 1 aims to develop single-image deblurring methods focusing on restoration quality.  ...  The winning methods demonstrate the state-ofthe-art performance on image and video deblurring tasks.  ...  CVML team -Track 1 CVML team uses a Wasserstein autoencoder [74] for single image deblurring. The latent space is represented as spatial tensor instead of 1D vector.  ... 
doi:10.1109/cvprw50498.2020.00216 dblp:conf/cvpr/NahSTLTXCTBSXCS20 fatcat:a6ojyfuidrbb3avwdpv4mje77e
« Previous Showing results 1 — 15 out of 122 results