摘要
大多数现有的客观图像质量评价算法往往针对单一失真类型设计,对混合多失真图像质量评价效果欠佳,而且大都是运用传统机器学习方法,很少用到深度学习方法,为此,提出一种基于相位一致变换和卷积神经网络的无参考图像质量评价方法,用来评价混合失真图像。对输入图像进行分块和相位一致变换,使用卷积网络训练、预测图像质量得分,其中卷积网络结构包括4层卷积层、3层最大池化层和2层全连接层。在Live混合失真质量评价数据库上的实验结果表明,所提方法预测的图像质量分和主观质量评分达到了很好的一致性。
Most of the existing algorithms of objective image quality assessment are only designed for a single type of distortion and they are not applicable for mixed distortion.Among these methods,most are traditional machine learning and few are deep learning.A no-reference image quality assessment method based on phase congruency and convolutional neural networks is proposed to evaluate the mixed distorted images.The input images are divided into patches and processed by phase congruency transformation,convolutional neural networks are used to train and predict quality scores of images.The convolutional network structure consists of 4 layers of convolutions,3 layers of maximum pooling,and 2 fully connected layers.The experimental results on the Live multiply distortion quality evaluation database show that the proposed method has a good consistency between the image quality and the subjective quality score.
作者
武利秀
桑庆兵
WU Lixiu;SANG Qingbing(School of Internet of Things Engineering,Jiangnan University,Wuxi 214122,China)
出处
《光学技术》
CAS
CSCD
北大核心
2018年第5期555-561,共7页
Optical Technique
基金
国家自然科学基金(61673194,61672265)
江苏省产学研前瞻性联合研究项目(BY2016022-17/001)
江苏省自然科学基金项目(BK20171142)
关键词
无参考
图像质量评价
卷积神经网络
相位一致性
深度学习
no reference
image quality evaluation
convolution neural networks
phase congruency
deep learning
