摘要
该文提出的评价方法通过相关系数矩阵的特征向量将评价指标线性变化成彼此独立的主成分,根据主成分累计贡献值确定主成分的取用维数,由主成分方差确定权重。其优点是:可以消除由于指标间的相关性带来的偏差,降低计算维数,从而降低指标选择的难度,提高评价结果的可信度;此外,可以消除人为确定指标权重引起的弊病,使评价结果更具客观性和准确性。
The evaluation method proposed in this paper changes from linear evaluation into the principal components which are independent of each other, the dimension of principal components are determined ac- cording to the cumulative contribution of principal components, and the weights are determined by the variance of component weights. The advantage is that proposed method can eliminate the bias caused by correlation be- tween indicators, reduce the computational dimension. Thereby the difficulty of the selection of indicators is reduced, and the credibility of evaluation results is enhanced. In addition, proposed method can eliminate the defects that are induced by artificially determine the index weight. And the evaluation results will be more ob- jective and accurate.
出处
《杭州电子科技大学学报(自然科学版)》
2012年第3期41-44,共4页
Journal of Hangzhou Dianzi University:Natural Sciences
基金
浙江省自然科学基金资助项目(Y1111229)
关键词
主成分分析法
多元线性回归
图像质量评价
相关系数
principal component analysis
multiple linear regression
image quality evaluation
correlation co-efficient