摘要
提出一种基于梯度特征的图像质量评价算法(GSIM)。该算法通过比较原始图像和降质图像的亮度、对比度和清晰度信息,得到归一化的图像质量评价指标。对Live数据库不同类型降质图像和实验采集的激光干扰图像的评价结果表明:与PSNR相比,GSIM算法的相关性方面提高了8.5%,准确率提高了7.6%,一致性提高了4.8%;与SSIM算法相比,GSIM算法的相关性方面提高了1.5%,准确率提高了3.8%,一致性提高了2.6%。证明了提出的GSIM算法相对MSE、PSNR及SSIM算法能准确评价交叉失真图像质量,评价结果更符合人的主观视觉感受。在评价基于掩盖效应的激光干扰图像时,GSIM算法能克服背景强度不一致带来的影响,评价结果能准确反映激光干扰效果。
This paper proposes a new algorithm based on gradient similarity （ GSIM ） .By comparing the original images with the disturbed ones on their luminance,contrast and resolution,an image quality assessment index system was obtained,and a modified GSIM was constructed .Comparison experiments show that the correlation,prediction accuracy and consistency of the proposed metric are respectively 8 .5%,7.6%and 4.8%higher than the PSNR index,and are respectively 1.5%,3.8% and 2.6% higher than the SSIM index .In terms of experiment results , the new algorithm shows better feasibility comparing with MSE,PSNR and SSIM image quality assessment methods .GSIM can overcome the influence of the difference in background intensity,and it is efficacious to reflect the laser-dazzling effect .
出处
《电光与控制》
北大核心
2014年第8期20-23,共4页
Electronics Optics & Control
关键词
激光干扰
图像质量评价
梯度相似度
laser-dazzling
image quality assessment
gradient similarity
