摘要
针对现有图像质量评估方法不能有效评价所有失真类型图像、计算量大等问题进行了研究,受人眼视觉特性的启发,提出了一种新的基于图像边缘结构相似性的全参考质量评价方法。该方法利用图像边缘梯度对图像失真的敏感性不同,通过Prewitt算子计算水平和竖直方向梯度向量绝对值和,进而定义了两种基于边缘结构相似性的图像质量评价标准:图像梯度结构相似度均值(GSSM)和图像梯度结构相似度标准差(GSSD)。标准图像数据库上的实验结果表明,GSSD方法优于GSSM和现有的全参考图象质量评价方法,主客观分值一致性好,并且计算简单,适合应用于大规模的图像质量评价。
Study for the existing image quality assessment methods can not effectively evaluate all types of image distortion, and large amount of calculation problems. This paper proposed a new full reference image quality assessment method based on edge structure similarity inspired by the human visual system. Considering the image gradient sensitivity was differrent when ob- serving different distortion image,the method computed the sum of horizontal gradient absolute value and vertical gradient abso- lute value by Prewitt filter operator,then proposed two image quality evaluation criteria based on edge structure similarity:im- age gradient structure similarity mean (GSSM) and image gradient structure similarity standard deviation (GSSD). Experimen- tal results on standard image database show that the GSSD methods is superior to GSSM melhod and existing full reference im- age quality assessment methods. It obtains high con'elations with subjective quality evaluations and low calculalion, and is more suitable for large-scale image quality assessment.
出处
《计算机应用研究》
CSCD
北大核心
2015年第9期2870-2873,共4页
Application Research of Computers
基金
中国科学院国防科技创新基金资助项目(CXJJ-14-Z65)
关键词
图像质量评价
全参考
PREWITT算子
梯度结构相似度均值
梯度结构相似度标准差
image quality assessment
full reference
Prewitt operator
gradient structure similarity mean
gradient structure- similarity standard deviation