BibTeX
@MISC{Jia_generalizedface,
author = {Kui Jia and Shaogang Gong},
title = {Generalized Face Super-Resolution},
year = {}
}
Share
 |
 |
 |
 |
||
OpenURL
Abstract
Abstract—Existing learning-based face super-resolution (hallucination) techniques generate high-resolution images of a single facial modality (i.e., at a fixed expression, pose and illumination) given one or set of low-resolution face images as probe. Here, we present a generalized approach based on a hierarchical tensor (multilinear) space representation for hallucinating high-resolution face images across multiple modalities, achieving generalization to variations in expression and pose. In particular, we formulate a unified tensor which can be reduced to two parts: a global image-based tensor for modeling the mappings among different facial modalities, and a local patch-based multiresolution tensor for incorporating high-resolution image details. For realistic hallucination of unregistered low-resolution faces contained in raw images, we develop an automatic face alignment algorithm capable of pixel-wise alignment by iteratively warping the probing face to its projection in the space of training face images. Our experiments show not only performance superiority over existing benchmark face super-resolution techniques on single modal face hallucination, but also novelty of our approach in coping with multimodal hallucination and its robustness in automatic alignment under practical imaging conditions. Index Terms—Face hallucination, super-resolution, tensor. I. PROBLEM STATEMENT
Keyphrases
multiple modality low-resolution face image high-resolution face image global image-based tensor space representation single modal face hallucination benchmark face super-resolution technique learning-based face super-resolution pixel-wise alignment high-resolution image detail index term face hallucination generalized approach different facial modality face image practical imaging condition unified tensor local patch-based multiresolution tensor problem statement fixed expression single facial modality unregistered low-resolution automatic face alignment algorithm high-resolution image performance superiority multimodal hallucination automatic alignment raw image realistic hallucination hierarchical tensor
