摘要
针对单一因素指标对图像融合质量评价的局限性,提出了一种异源图像融合质量评价模型。首先,计算出各融合图像的多种典型图像融合客观评价值,利用离散化将每项指标分为三类;然后运用粗糙集对进行约简,消除冗余以及冲突指标,将简化的指标作为BP神经网络的输入样本,将融合图像的主观评价结论样本作为模糊期望输出,通过学习生成网络;最后输入训练样本,得到输出的主观评价指标。大量实验结果表明该方法评价结果合理,主客观评价有较好的一致性,为融合图像自动化评价的实现提供了有效的途径。
This paper proposed an evaluation model for different source image fusion quality to resolve the limitations of single factor index on the quality evaluation. Firstly, the model calculated a variety of typical objective evaluation indexes of the fusion images and each index was divided into three categories by discretization, then reduced the indexs by the use of the rough set and eliminated the redundancy and conflict, then inputted the residual indexes to BP neural network and outputted the subjective evaluation conclusion sample and the neural network was generated. Finally, the model inputted the test samples to the trained BP neural network and got the subjective evaluation conclusion. Series of experimental results show that the method is reasonable and consistent with subjective evaluation and provides a valuable way for the realization of automatic fusion image evaluation.
出处
《红外技术》
CSCD
北大核心
2011年第10期568-573,共6页
Infrared Technology
基金
航空科学基金资助项目
编号:20090153002
关键词
图像融合
综合质量评价
粗糙集
BP神经网络
image fusion, comprehensive quality evaluation, rough set, BP neural network