Image up-sampling using total-variation regularization with a new observation model

By Hussein H. Aly, Eric Dubois. 2005

362 citations recorded by the source corpus.

Subfields: Computer Vision and Pattern Recognition

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

Citations in the corpus

Photo-Realistic Single Image Super-Resolution Using a Generative Adversarial Network (2016, methodology): “We further employ a regularizer based on the total variation to encourage spatially coherent solutions [2, 26].”
AutoGAN-Distiller: Searching to Compress Generative Adversarial Networks (2020, positive): “…use a combination of content loss L c (avoid color shift) and perceptual loss L p (preserve visual and semantic details) as defined in (Johnson et al., 2016), in addition to a total variation loss L tv (enforce pixel-level similarity) (Aly & Dubois, 2005): where β 1 , β 2 , β 3 are hyperparameters.”
Fast Patch-based Style Transfer of Arbitrary Style (2016, methodology): “where we’ll say that the synthesized image is of dimension h by w by d, || · ||F is the Frobenius norm, and `TV (·) is the total variation regularization term widely used in image generation methods [1, 17, 26].”
StyleBank: An Explicit Representation for Neural Image Style Transfer (2017, background): “Ltv(Oi) is a variation regularizer used in [2, 19].”
Deep Learning for Image Super-Resolution: A Survey (2019, methodology): “In order to suppress noise in generated images, the total variation (TV) loss [135] is introduced into the SR ﬁeld by Aly et al. [136].”
Image super-resolution using gradient profile prior (2008, methodology): “Generic smoothness prior [25, 11] and edge smoothness prior [21, 1, 6, 7, 22, 27] are two widely used priors.”
Deeply Improved Sparse Coding for Image Super-Resolution (2015, background): “Analytical priors, such as bicubic interpolation, work well for smooth regions; while image models based on statistics of edges [9] and gradients [15, 1] can recover sharper structures.”
Recovering Realistic Texture in Image Super-Resolution by Deep Spatial Feature Transform (2018, methodology): “Image priors such as edge features [13, 41], statistics [24, 1] and internal patch recurrence [16] are employed to improve performance.”
Single-Image Super-Resolution: A Benchmark (2014, methodology): “Total variation has also been used as a regularization term for generating HR images [1,48].”
DSLR-Quality Photos on Mobile Devices with Deep Convolutional Networks (2017, background): “(6) where C j, H j and W j denotes the number, height and width of the feature maps, F W(I s) — enhanced image. 3.1.4 Total variation loss. In addition to previous losses, we add total variation loss [7] to enforce spatial smoothness of the produced images. The loss is deﬁned as: L tv = 1 CHW kr xF ”
Image upsampling via imposed edge statistics (2007, methodology): “This approach has been formulated for image upsampling in [Aly and Dubois 2005].”
Super-Resolution With Sparse Mixing Estimators (2010, methodology): “More global image models impose image smoothness priors suc h as a bounded total variation to optimize the interpolation [39], [44], [3], [43].”
Accurate Blur Models vs. Image Priors in Single Image Super-resolution (2013, background): “Developing sophisticated image priors has been the focus of much single image SR research in the past decade, with many significant successes [7, 20, 23, 8, 6, 9, 21, 27, 25, 5, 19, 22, 1, 8, 13].”
Single image super-resolution using Gaussian process regression (2011, background): “, [1, 5, 19]) and impose the constraint that when properly downsampled, the HR image should reproduce the original LR image.”
Deep Networks for Image Super-Resolution with Sparse Prior (2015, background): “Analytical priors, such as bicubic interpolation, work well for smooth regions; while image models based on statistics of edges [11] and gradients [17, 1] can recover sharper structures.”