Blind dehazing using internal patch recurrence

Bahat, Yuval; Irani, Michal

doi:10.1109/iccphot.2016.7492870

Cited by 80 publications

(72 citation statements)

References 16 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…Modeling the internal distribution of patches within a single natural image has been long recognized as a powerful prior in many computer vision tasks [62]. Classical examples include denoising [63], deblurring [39], super resolution [17], dehazing [2,14], and image editing [47,37,20,9]. Motivated by these works, here we show how SinGAN can be used within a simple unified framework to solve a variety of image manipulation tasks, including paint-toimage, editing, harmonization, super-resolution, and anima- Figure 2: Image manipulation.…”

Section: Introductionmentioning

confidence: 99%

SinGAN: Learning a Generative Model From a Single Natural Image

Shaham

Dekel

Michaeli

2019

2019 IEEE/CVF International Conference on Computer Vision (ICCV)

754

689

View full text Add to dashboard Cite

Random samples from a single image Single training image Figure 1: Image generation learned from a single training Image. We propose SinGAN-a new unconditional generative model trained on a single natural image. Our model learns the image's patch statistics across multiple scales, using a dedicated multi-scale adversarial training scheme; it can then be used to generate new realistic image samples that preserve the original patch distribution while creating new object configurations and structures. AbstractWe introduce SinGAN, an unconditional generative model that can be learned from a single natural image. Our model is trained to capture the internal distribution of patches within the image, and is then able to generate high quality, diverse samples that carry the same visual content as the image. SinGAN contains a pyramid of fully convolutional GANs, each responsible for learning the patch distribution at a different scale of the image. This allows generating new samples of arbitrary size and aspect ratio, that have significant variability, yet maintain both the global structure and the fine textures of the training image. In contrast to previous single image GAN schemes, our approach is not limited to texture images, and is not conditional (i.e. it generates samples from noise). User studies confirm that the generated samples are commonly confused to be real images. We illustrate the utility of SinGAN in a wide range of image manipulation tasks.

show abstract

Section: Introductionmentioning

confidence: 99%

SinGAN: Learning a Generative Model From a Single Natural Image

Shaham

Dekel

Michaeli

2019

2019 IEEE/CVF International Conference on Computer Vision (ICCV)

754

689

View full text Add to dashboard Cite

show abstract

“…Handling non-uniform airlight: Most single-image blind dehazing methods (e.g. [23,28,18,6,8]) assume a uniform airlight color A for the entire image (i.e., A(x) ≡ A). This is true also for deep network based dehazing methods [10,30], which train on synthesized datasets of hazy/non-hazy image pairs.…”

Section: Image Dehazingmentioning

confidence: 99%

“…While it satisfies the strong internal self-similarity requirement, it tends to be much smoother than a natural image, and should not deviate much from a global airlight color. Hence, we apply an extra regularization loss on the airlight layer: A(x) − A 2 , where A is a single initial airlight color estimated from the hazy image I using one of the standard methods (we used the method of [6]). Although the deviations from the initial airlight A are subtle, they are quite crucial to the quality of the recovered haze-free image (see Fig.…”

Section: Image Dehazingmentioning

confidence: 99%

“Double-DIP”: Unsupervised Image Decomposition via Coupled Deep-Image-Priors

Gandelsman

Shocher

Irani

2019

2019 IEEE/CVF Conference on Computer Vision and Pattern Recognition (CVPR)

Self Cite

274

194

View full text Add to dashboard Cite

Figure 1: A unified framework for image decomposition. An image can be viewed as a mixture of "simpler" layers. Decomposing an image into such layers provides a unified framework for many seemingly unrelated vision tasks (e.g., segmentation, dehazing, transparency separation). Such a decomposition can be achieved using "Double-DIP". AbstractMany seemingly unrelated computer vision tasks can be viewed as a special case of image decomposition into separate layers. For example, image segmentation (separation into foreground and background layers); transparent layer separation (into reflection and transmission layers); Image dehazing (separation into a clear image and a haze map), and more. In this paper we propose a unified framework for unsupervised layer decomposition of a single image, based on coupled "Deep-image-Prior" (DIP) networks. It was shown [38] that the structure of a single DIP generator network is sufficient to capture the low-level statistics of a single image. We show that coupling multiple such DIPs provides a powerful tool for decomposing images into their basic components, for a wide variety of applications. This capability stems from the fact that the internal statistics of a mixture of layers is more complex than the statistics of each of its individual components. We show the power of this approach for Image-Dehazing, Fg/Bg Segmentation, Watermark-Removal, Transparency Separation in images and video, and more. These capabilities are achieved in a totally unsupervised way, with no training examples other than the input image/video itself. 1

show abstract

“…Kernel k is often assumed to be sparse [1][2][3] and continuous [3,4]. For I , prior knowledge about natural images has been widely exploited to construct different regularizations.…”

Section: Introductionmentioning

confidence: 99%

“…Recently, image patches have been used for image deblurring [3,14,15], which can utilize more pixels in the neighborhood of the interest area and can better adapt to different local features. Sun et al [14] established an image patch prior subset tailored toward image edge and corner primitives with external clear images.…”

Section: Introductionmentioning

confidence: 99%

Motion blur image deblurring using edge-based color patches

Zhao¹,

Kan²

2019

Turk J Elec Eng & Comp Sci

View full text Add to dashboard Cite

The shaking of a camera can easily cause blurs in an image. Thus, deblurring is a problem that is worth solving and has always been an active research interest. The color information in an image is an important feature and contains clues for image deblurring that have not been widely exploited. In this paper, we present an efficient and stable blurring kernel estimation method by solving an energy function constructed by a weighted color approximation regularization term. The term is derived from a two-color model, and we use a defined weight to alleviate the color change through the blurring process. Then we select salient edges in an effective way to apply the proposed method on the patches centered at these edges. Experiments on synthetic and real-world images show the efficiency and stability of our proposed method.

show abstract

Blind dehazing using internal patch recurrence

Cited by 80 publications

References 16 publications

SinGAN: Learning a Generative Model From a Single Natural Image

SinGAN: Learning a Generative Model From a Single Natural Image

“Double-DIP”: Unsupervised Image Decomposition via Coupled Deep-Image-Priors

Motion blur image deblurring using edge-based color patches

Contact Info

Product

Resources

About