PRICAI 2016 Paper Selected | Large-Scale Coupling Mapping for Low-Resolution Face Recognition

Large Margin Coupled Mapping for Low Resolution Face Recognition

Summary: Traditional face recognition algorithms can achieve very high performance in a well-controlled environment. However, when the resolution of facial images changes, the performance of these algorithms is very low. A two-step framework was proposed to solve the resolution problem by using super-resolution (SR) and face recognition on super-resolution face images. However, when SR focuses more on visual enhancement than on classification accuracy, the method's performance in identifying tasks is usually lower. Recently, Collage Mapping (CM) with different resolutions has been introduced into the face recognition framework, which learns the common feature subspaces of a high-resolution (HR) and low-resolution (LR) face image. In this paper, inspired by the maximum edge projection, we propose a large-scale coupled mapping (LMCM) algorithm to learn to predict to maximize the inter-class distance between objects and in public spaces. Experimental results from the public FERET and SCface databases show that LMCM is effective for low-resolution face recognition.

Keywords: Coupled mapping , low-resolution face recognition , large-amplitude coupling mapping , FERET SCface

First author introduction

Jiaqi Zhang

Position: Graduate School of Shenzhen, Tsinghua University

Research Interests: Artificial Intelligence, Pattern Recognition, Computer Vision, Image Processing

Related academic research

· "Off-line Signature Verification using Local Features and Decision Trees" (International Journal of Pattern Recognition and Artificial Intelligence 2016)

· "Dynamic Background Estimation and Complementary Learning for Pixel-wise Foreground/Background Segmentation"

Via:PRICAI 2016

PS : This article was compiled by Lei Feng Network (search "Lei Feng Network" public number) and it was compiled without permission.

Original paper download