PULSE: Self-Supervised Photo Upsampling via Latent Space Exploration of Generative Models

• URL: http://arxiv.org/abs/2003.03808v3
• Date: Mon, 20 Jul 2020 21:38:32 GMT
• Title: PULSE: Self-Supervised Photo Upsampling via Latent Space Exploration of Generative Models
• Authors: Sachit Menon, Alexandru Damian, Shijia Hu, Nikhil Ravi, Cynthia Rudin
• Abstract summary: PULSE (Photo Upsampling via Latent Space Exploration) generates high-resolution, realistic images at resolutions previously unseen in the literature. Our method outperforms state-of-the-art methods in perceptual quality at higher resolutions and scale factors than previously possible.
• Score: 77.32079593577821
• License: http://arxiv.org/licenses/nonexclusive-distrib/1.0/
• Abstract: The primary aim of single-image super-resolution is to construct high-resolution (HR) images from corresponding low-resolution (LR) inputs. In previous approaches, which have generally been supervised, the training objective typically measures a pixel-wise average distance between the super-resolved (SR) and HR images. Optimizing such metrics often leads to blurring, especially in high variance (detailed) regions. We propose an alternative formulation of the super-resolution problem based on creating realistic SR images that downscale correctly. We present an algorithm addressing this problem, PULSE (Photo Upsampling via Latent Space Exploration), which generates high-resolution, realistic images at resolutions previously unseen in the literature. It accomplishes this in an entirely self-supervised fashion and is not confined to a specific degradation operator used during training, unlike previous methods (which require supervised training on databases of LR-HR image pairs). Instead of starting with the LR image and slowly adding detail, PULSE traverses the high-resolution natural image manifold, searching for images that downscale to the original LR image. This is formalized through the "downscaling loss," which guides exploration through the latent space of a generative model. By leveraging properties of high-dimensional Gaussians, we restrict the search space to guarantee realistic outputs. PULSE thereby generates super-resolved images that both are realistic and downscale correctly. We show proof of concept of our approach in the domain of face super-resolution (i.e., face hallucination). We also present a discussion of the limitations and biases of the method as currently implemented with an accompanying model card with relevant metrics. Our method outperforms state-of-the-art methods in perceptual quality at higher resolutions and scale factors than previously possible.

Related papers

• Realistic Extreme Image Rescaling via Generative Latent Space Learning [51.85790402171696]
  We propose a novel framework called Latent Space Based Image Rescaling (LSBIR) for extreme image rescaling tasks. LSBIR effectively leverages powerful natural image priors learned by a pre-trained text-to-image diffusion model to generate realistic HR images. In the first stage, a pseudo-invertible encoder-decoder models the bidirectional mapping between the latent features of the HR image and the target-sized LR image. In the second stage, the reconstructed features from the first stage are refined by a pre-trained diffusion model to generate more faithful and visually pleasing details.
  arXiv  Detail & Related papers  (2024-08-17T09:51:42Z)
• Real Image Super-Resolution using GAN through modeling of LR and HR process [20.537597542144916]
  We propose a learnable adaptive sinusoidal nonlinearities incorporated in LR and SR models by directly learn degradation distributions. We demonstrate the effectiveness of our proposed approach in quantitative and qualitative experiments.
  arXiv  Detail & Related papers  (2022-10-19T09:23:37Z)
• Memory-augmented Deep Unfolding Network for Guided Image Super-resolution [67.83489239124557]
  Guided image super-resolution (GISR) aims to obtain a high-resolution (HR) target image by enhancing the spatial resolution of a low-resolution (LR) target image under the guidance of a HR image. Previous model-based methods mainly takes the entire image as a whole, and assume the prior distribution between the HR target image and the HR guidance image. We propose a maximal a posterior (MAP) estimation model for GISR with two types of prior on the HR target image.
  arXiv  Detail & Related papers  (2022-02-12T15:37:13Z)
• Toward Real-world Image Super-resolution via Hardware-based Adaptive Degradation Models [3.9037347042028254]
  Most single image super-resolution (SR) methods are developed on synthetic low-resolution (LR) and high-resolution (HR) image pairs. We propose a novel supervised method to simulate an unknown degradation process with the inclusion of prior hardware knowledge. Experiments on the real-world datasets validate that our degradation model can estimate LR images more accurately than the predetermined degradation operation.
  arXiv  Detail & Related papers  (2021-10-20T19:53:48Z)
• Best-Buddy GANs for Highly Detailed Image Super-Resolution [71.13466303340192]
  We consider the single image super-resolution (SISR) problem, where a high-resolution (HR) image is generated based on a low-resolution (LR) input. Most methods along this line rely on a predefined single-LR-single-HR mapping, which is not flexible enough for the SISR task. We propose best-buddy GANs (Beby-GAN) for rich-detail SISR. Relaxing the immutable one-to-one constraint, we allow the estimated patches to dynamically seek the best supervision.
  arXiv  Detail & Related papers  (2021-03-29T02:58:27Z)
• Real-World Super-Resolution of Face-Images from Surveillance Cameras [25.258587196435464]
  We propose a novel framework for generation of realistic LR/HR training pairs. Our framework estimates realistic blur kernels, noise distributions, and JPEG compression artifacts to generate LR images with similar image characteristics as the ones in the source domain. For better perceptual quality we use a Generative Adrial Network (GAN) based SR model where we have exchanged the commonly used VGG-loss [24] with LPIPS-loss [52]
  arXiv  Detail & Related papers  (2021-02-05T11:38:30Z)
• SRFlow: Learning the Super-Resolution Space with Normalizing Flow [176.07982398988747]
  Super-resolution is an ill-posed problem, since it allows for multiple predictions for a given low-resolution image. We propose SRFlow: a normalizing flow based super-resolution method capable of learning the conditional distribution of the output. Our model is trained in a principled manner using a single loss, namely the negative log-likelihood.
  arXiv  Detail & Related papers  (2020-06-25T06:34:04Z)
• Deep Generative Adversarial Residual Convolutional Networks for Real-World Super-Resolution [31.934084942626257]
  We propose a deep Super-Resolution Residual Convolutional Generative Adversarial Network (SRResCGAN) It follows the real-world degradation settings by adversarial training the model with pixel-wise supervision in the HR domain from its generated LR counterpart. The proposed network exploits the residual learning by minimizing the energy-based objective function with powerful image regularization and convex optimization techniques.
  arXiv  Detail & Related papers  (2020-05-03T00:12:38Z)
• Characteristic Regularisation for Super-Resolving Face Images [81.84939112201377]
  Existing facial image super-resolution (SR) methods focus mostly on improving artificially down-sampled low-resolution (LR) imagery. Previous unsupervised domain adaptation (UDA) methods address this issue by training a model using unpaired genuine LR and HR data. This renders the model overstretched with two tasks: consistifying the visual characteristics and enhancing the image resolution. We formulate a method that joins the advantages of conventional SR and UDA models.
  arXiv  Detail & Related papers  (2019-12-30T16:27:24Z)

This list is automatically generated from the titles and abstracts of the papers in this site.

This site does not guarantee the quality of this site (including all information) and is not responsible for any consequences.