基于模糊度和噪声水平的图像质量评价方法  

Image quality assessment based on blurring and noise level

作　　者：赵巨峰[1] 冯华君[1] 徐之海[1] 李奇[1] 

机构地区：[1]浙江大学现代光学仪器国家重点实验室,浙江杭州310027

出　　处：《光电子．激光》2010年第7期1062-1066,共5页Journal of Optoelectronics·Laser

基　　金：国家"973"计划资助项目(2009CB724006);国家自然科学基金资助项目(60977010);国家"863"计划资助项目(2009AA12Z108)

摘　　要：针对图像质量评价的重要性,提出了一种新的无参考图像质量客观评价方式。算法考虑了模糊度与噪声水平两方面,用平均边缘宽度衡量图像的模糊度,通过比较去噪前后的图像预测图像受噪声污染的程度,最后通过两者的综合作为无参考图像质量评价指标。实验结果表明:将模糊度和噪声评价相结合,具有很强的抗噪性和广泛的适用范围;与峰值信噪比(PSNR)和结构类似性(SSIM)等算法比较,本文算法可以很好地区分各种失真类型图像的质量好坏,其结果接近人眼的主观感受。The quality assessment of objective images is important because of its valuable application.A novel no-reference objective image quality assessment is presented based on blurring and noise level.Blurring is evaluated by computing the average width of the edge of image.Meanwhile we can compare the differences between distortion image and its denoised image to predict the noise level.The assessment result is given by combining the blurring with noise level and balancing their different sensitivities to different image distortions.With considering blurring and noise level together,this method has a wide application and excellent anti-noise ability.This assessment is particularly experimented,which proves it more effective comparing with other methods such as PSNR and SSIM.

关 键 词：无参考 图像质量评价 边缘宽度 噪声水平 

分 类 号：TP391[自动化与计算机技术—计算机应用技术]