摘要
分析了传统的基于灰度误差的图像质量评价算法的局限,扩展了图像质量评价的定义.考虑原始图像和畸变图像尺寸不一致的情况,提出了用奇异值分解把图像矩阵转为向量,用奇异值向量之间的夹角作为图像的质量评价指标.实验表明:提出的基于奇异值向量夹角的准则对压缩、噪声和旋转、平移、尺度等几何畸变等都具有好的性质;且适于文中扩展的质量评价的定义.最后对实验结果和人类视觉系统的主观评价进行了比较分析.
The limitations of several classic image quality measure algorithms based on error of grayscale are analyzed, and the definition of image quality measure is extended, according to which the size of the original image can be different from that of the distorted image. The image matrix is transformed into vector by singular value decomposition. The angle between singular vectors of the original image and the distorted image is used to measure the image quality. The experimental results show that the algorithm proposed has good properties to image compression, noise and geometry distortion including scale, translation and rotation transform, and can be applied to the image quality measure definition proposed in this paper. The results are analyzed and compared with the measurement of human visual system.
出处
《东南大学学报（自然科学版）》
EI
CAS
CSCD
北大核心
2006年第4期643-646,共4页
Journal of Southeast University：Natural Science Edition
基金
国家自然科学基金资助项目(50275078)
关键词
图像质量评价
奇异值分解
人类视觉系统
image quality measure
singular value decomposition
human visual system
