TY - JOUR
T1 - Low-resolution image categorization via heterogeneous domain adaptation
AU - Yao, Yuan
AU - Li, Xutao
AU - Ye, Yunming
AU - Liu, Feng
AU - Ng, Michael K.
AU - Huang, Zhichao
AU - Zhang, Yu
N1 - Funding Information:
This work was supported by the National Key R&D Program of China, 2018YFB0504900, 2018YFB0504905 and the Shenzhen Science and Technology Program, China under Grant JCYJ20170811160212033, and NSFC, China under Grant Nos. 61602132.
Publisher copyright:
© 2018 Elsevier B.V.
PY - 2019/1/1
Y1 - 2019/1/1
N2 - Most of existing image categorizations assume that the given datasets have a good resolution and quality. However, the assumption is often violated in real applications. In this paper, we study the low-resolution (LR) image categorization. By utilizing labeled high-resolution (HR) images as auxiliary information, we formulate the problem as a heterogeneous domain adaptation problem and propose a Discriminative Joint Distribution Adaptation (DJDA) model to solve it. The DJDA model projects both LR and HR images into an intermediate subspace with a well-designed objective function, where the distance between classes is expected to be enlarged and the distribution divergence to be reduced. As a result, the discriminative knowledge for HR images can be transferred effectively to LR images. Experimental results demonstrate the proposed DJDA method produces significantly superior categorization accuracies against state-of-the-art competitors.
AB - Most of existing image categorizations assume that the given datasets have a good resolution and quality. However, the assumption is often violated in real applications. In this paper, we study the low-resolution (LR) image categorization. By utilizing labeled high-resolution (HR) images as auxiliary information, we formulate the problem as a heterogeneous domain adaptation problem and propose a Discriminative Joint Distribution Adaptation (DJDA) model to solve it. The DJDA model projects both LR and HR images into an intermediate subspace with a well-designed objective function, where the distance between classes is expected to be enlarged and the distribution divergence to be reduced. As a result, the discriminative knowledge for HR images can be transferred effectively to LR images. Experimental results demonstrate the proposed DJDA method produces significantly superior categorization accuracies against state-of-the-art competitors.
KW - Heterogeneous domain adaptation
KW - Low-resolution image categorization
KW - Subspace learning
UR - http://www.scopus.com/inward/record.url?scp=85054003702&partnerID=8YFLogxK
U2 - 10.1016/j.knosys.2018.09.027
DO - 10.1016/j.knosys.2018.09.027
M3 - Journal article
AN - SCOPUS:85054003702
SN - 0950-7051
VL - 163
SP - 656
EP - 665
JO - Knowledge-Based Systems
JF - Knowledge-Based Systems
ER -