摘要
盲图像质量评价是指在没有原始图像信息的情况下,预测给定图像的视觉感知质量.迄今为止,基于无监督特征学习的盲图像质量评价方法取得了较好的性能,但其质量预测精度随特征维度的降低而显著下降.为了克服这一缺陷,作者将主动学习策略与无监督特征学习相结合,提出了一种主动特征学习框架,以提高图像特征表示的判别性,并利用所学特征进行质量预测.实验表明,在特征维度较低时,与基于无监督特征学习的方法相比,文中方法在图像质量预测精度上提高了8%.同时,文中方法的性能也优于现有的其他盲图像质量评价方法.
Blind image quality assessment(BIQA)aims to predict human perceived image quality without access to reference images.For now,the BIQA algorithms based on unsupervised feature learning have shown promising results,but their performance dramatically decreases as the dimension of the feature vector becomes lower.To combat this limitation,we propose an active feature learning framework which introduces the methodology of active learning into unsupervised feature learning in order to improve the discriminative power of the learned image representation.Afterwards,we utilize the learned image representation for quality prediction.Thorough experiments on the LIVE database demonstrate that when the feature vector is of low dimension,the proposed method outperforms the methods based on unsupervised feature learning by 8%.In addition,the performance of the proposed method is distinctly better than state-of-the-art BIQA methods.
出处
《计算机学报》
EI
CSCD
北大核心
2014年第10期2227-2234,共8页
Chinese Journal of Computers
基金
国家杰出青年科学基金(61125204)
国家自然科学基金(61172146
61372130)
中央高校基本科研业务费专项资金(K5051202048
JB149901
BDY081426
JB140214)
教育部"创新团队发展计划"(IRT13088)
陕西省重点科技创新团队(2012KCT-02)资助
关键词
主动学习
字典学习
特征学习
图像表示
图像质量评价
active learning
dictionary learning
feature learning
image representation
image quality assessment