Face Hallucination: Theory and Practice

By Ce Liu, Harry Shum, William T. Freeman. 2007

437 citations recorded by the source corpus.

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

In Media Technology

Citations in the corpus

Image Super-Resolution Using Deep Convolutional Networks (2014, background): “Index Terms—Super-resolution, deep convolutional neural networks, sparse coding F”
Single image super-resolution using Gaussian process regression (2011, background): “, [14, 7, 4, 22]), detailed textures are hallucinated by searching through a training set of LR/HR image or patch pairs.”
FSRNet: End-to-End Learning Face Super-Resolution with Facial Priors (2017, neutral): “, [24] needs 8 minutes and [14] needs 15− 20 minutes.”
Image Super-Resolution Via Sparse Representation (2010, methodology): “[16] proposed a two-step statistical approach integrating the global PCA model and a local patch model.”
Photorealistic Facial Texture Inference Using Deep Neural Networks (2016, background): “[29] hallucinates high-frequency content using a local path-based Markov network, but the results remain relatively blurry and cannot predict missing regions.”
Beyond pixels: exploring new representations and applications for motion analysis (2009, background): “Freeman, I designed the first practical face hallucination system [65] to detect, align and super-resolve small faces in photos.”
Expression flow for 3D-aware face component transfer (2011, methodology): “Earlier approaches use generic face images as training data for applications such as super-resolution [Liu et al. 2007] and attractiveness enhancement by global face warping [Leyvand et al.”
Ultra-Resolving Face Images by Discriminative Generative Networks (2016, methodology): “’s method [5]: Since this method requires the face images in the dataset to be precisely aligned, it is difficult for it to learn a representative subspace from the CelebA dataset where face images have large variations.”
Recover Canonical-View Faces in the Wild with Deep Neural Networks (2014, background): “The aim of face hallucination is to reconstruct high-resolution face images from low-resolution ones [26] , or to remove glasses from face images 口9].”
Super resolution using edge prior and single image detail synthesis (2010, background): “This is described briefly in Section 3.1 as necessary for implementation; further details can be found in [23].”
Image super-resolution as sparse representation of raw image patches (2008, background): “Another class of super-resolution methods that can overcome this difficulty are learning based approaches, which use a learned co-occurrence prior to predict the correspondence between low-resolution and high-resolution image patches [12, 26, 16, 5, 20].”
Global-Local Face Upsampling Network (2016, methodology): “Compared to the linear eigenface model of [1], the multi-layer nonlinear embedding and reconstruction used in our framework enables more effective encoding of characteristic facial features, in addition to variations such as alignment, face pose, and illumination.”
A Comprehensive Survey to Face Hallucination (2013, methodology): “Although the method (Liu et al. 2001, 2007a) is introduced in Sect.”
Image super-resolution using gradient profile prior (2008, methodology): “The learning based methods [10, 9, 26, 5, 28, 2, 7, 20, 31] “hallucinate” high frequency details from a training set of HR/LR image pairs.”
Very low resolution face recognition problem (2010, methodology): “Existing learning-based face SR algorithms can be roughly categorized into two approaches, namely examplebased approach [9] [11] [10] [12] [18] and maximum a posterior (MAP) based approach[1] [2] [8].”