摘要
考虑到在结构相似度(SSIM,structural similarity)模型中,亮度、对比度和结构度3个评价因子对不同失真类型图像质量评价(QA)的贡献程度不同,本文提出了根据图像失真类型分析的自适应SSIM(ASSIM)的IQA方法。首先,分析失真图像和参考图像的小波子带能量、傅里叶功率谱和幅度谱的数据特点,据此判定图像失真类型,包括高斯白噪声(WN)、JPEG压缩(JPEM)、高斯模糊(Gblur)及类JP2K4类失真;接着,通过优化算法确定SSIM在评价不同失真类型图像时最佳的评价因子权重;最后,将图像的失真类型判别和评价因子的调整相结合,实现对图像的自适应评价。实验结果表明,由于失真类型的判断和评价因子权值的优化,ASSIM对各类失真图像的评价效果都要优于SSIM,特别是对Gblur失真的图像进行评价时,Pearson系数(CC)值提高了0.05,Spearman等级相关系数(SROCC)值的提高超过0.039。
Considering the fact that three factors,that is,light,contrast and structure in structural similarity （SSIM） metric,have different contributions to image evaluation with different types of distortion,this paper proposes an adaptive image quality assessment,called as ASSIM,based on analysis of image’ s distortion types.Firstly,we analyze the characteristics of the original and distorted images,including the e nergy of wavelet sub-bands,power spectrum and magnitude spectrum of Fourier transform.Then the distortion type of an image is identified according to these characteristics,including Gaussian noise distortion,JPEG compression dis tortion,Gaussian blurring distortion and JPEG2000compression distortion.In accordance with the distortion type,the most suitable parameters, which indicate the weight distribution of the three factors in SSIM metric,are selected through the optimization.Finally,based on these parameters,the ASSIM a pproach is realized.The experimental results show that the proposed ASSIM outperforms SSIM,because the distortion type of i mage has been identified ahead and thus we can choose the most suitable parameters in the evaluation metric,par ticularly in the case of Gaussian blurring distortion,where Pearson value is increased by 0.05and Spearman value is improved at least 0.039,OR value and RMSE value are decreased by 0.1412and 1.6919,respectively.
出处
《光电子．激光》
EI
CAS
CSCD
北大核心
2014年第2期378-385,共8页
Journal of Optoelectronics·Laser
基金
国家自然科学基金(U1301257
61171163
61271270
61271021
61311140262)资助项目
