摘要
为了能够在图像质量评价领域实现自监督学习,提出一种基于半监督学习的双分支网络训练的无参考图像质量评价算法。它是具有两个分支的训练过程,其中在一个分支使用少量手工标记数据样本来进行有监督学习,在另一个分支进行自监督学习来辅助前者训练同一个特征提取器,自监督学习部分采用几种传统的全参考方法联合为训练样本打上软标签。在6个公开的图像数据库中进行大量实验,结果表明所提算法不仅在合成失真图像数据集上优于目前大多数方法,而且在真实失真图像数据集上具有良好的泛化性能,预测结果与人类主观感知表现一致。
This paper proposes a noreference image quality evaluation algorithm based on semisupervised learning and dualbranch network training to realize selfsupervised learning in image quality evaluation.Specifically,it is a training process with two branches in which a small number of handlabeled data samples are used for supervised learning in one branch.Selfsupervised learning is performed in the other branch to assist the former in training the same feature extractor;the selfsupervised learning part adopts several traditional fullreference methods to jointly label the training samples with soft labels.Extensive experiments are conducted on six public image databases.The results show that the proposed algorithm outperforms most current methods on the synthetic distorted image datasets and has a good generalization performance on the real distorted image datasets.The predicted results of the proposed algorithm are consistent with human subjective perception performance.
作者
金向东
桑庆兵
Jin Xiangdong;Sang Qingbing(School of Artificial Intelligence and Computer Science,Jiangnan University,Wuxi 214122,Jiangsu,China)
出处
《激光与光电子学进展》
CSCD
北大核心
2023年第4期262-269,共8页
Laser & Optoelectronics Progress
关键词
图像质量评价
特征提取
自监督学习
无参考
联合训练
image quality evaluation
feature extraction
selfsupervised learning
no reference
joint training
作者简介
通信作者:桑庆兵,sangqb@163.com。
