Empirical studies showed that between and resolution is required for face recognition.
Two SR approaches: MAP and example based. But similarity in LR image space doesn't imply similarity in HR image space. RLSR and DSR are proposed to solver VLR problem.
Evaluated these methods with reconstructed SR image with 1NN / SVM.
Influence of low resolution of images on reliability of face detection and recognition
Analyzed minimum requirements for the resolution of facial images. They used Haar and Eigenface, which seem out of date (This is a 2015 paper anyhow).
Face Hallucination and Recognition
Use an eigentransformation based hallucination method to improve the image resolution and it's helpful for recognition.
Face Hallucination: How Much It Can Improve Face Recognition
hen resolution is below it's hard to recognize.
In extramely low dimension, some of the face hallucination approaches do not work properly.
PSNR and RMSE values are not able to exactly reflect the hallucinating performance in terms of assisting recognition.
Face hallucination and recognition in social network services
Proposed a learning based face hallucination approach.
Recognition at a long distance: very low resolution face recognition and hallucination
2 ways: 1) SR; 2) train on LR images. Proposed joint approach.
Global Face reconstruction for face hallucination using orthogonal canonical correlation analysis
Skip. A global face reconstruction framework for face hallucination based on CCA.
Face hallucination with pose variation
Skip. Presented a framework for face hallucination with pose variation.
Structured face hallucination
Skip.
Facial image super resolution using sparse representation for improving face recognition in surveillance monitoring
Super resolve the image using sparse representation with the specific dictionary involving many natural and facial images followed by HMM and SVM based face recognition.
The MegaFace Benchmark: 1 million faces for recognition at scale
FR experiments on a LFW show stunning performance. This paper advocated evaluations at the million scale.
SCface – surveillance cameras face database
Dataset.
Large Scale Unconstrained Open Set Face Database
UCCS dataset.
Recognition at a long distance
Joint Face hallucination and Face recognition, conventional method.
DeepFace: Closing the Gap to Human-Level Performance in Face Verification
DNN 对人脸抽取特征，explicit 3d modeling 获取 alignment。
Deep learning face representation by joint identification-verification
DeepID。
Learning face representation from scratch
半自动收集 CASIA。突出数据越多越正义的真理。
FaceNet: A Unified Embedding for Face Recognition and Clustering
不学习中间特征，直接最优化图像之间距离。
The CMU Pose, Illumination, and Expression Database
用了高端照相设备收集数据 PIE 数据集，发了 PAMI。
The FERET Evaluation Methodology for Face-Recognition Algorithms
解决了数据集和评估系统的问题。
Labeled Faces in the Wild: A Database for Studying Face Recognition in Unconstrained Environments
