摘要
本文提出一种新的图像融合客观评价方法.首先,利用模糊C-均值聚类算法(FCM)对图像进行区域分割,获取区域特征矩阵.通过降维处理,计算区域的距离,以此作为区域相似度.其次,以对应区域的特征向量元素相似比作为对应区域的权重.区域的边缘强度比值作为区域之间的权重.最终获得图像的相似度.实验结果表明这种相似性度量考虑了图像像素的局部关系以及区域的显著性,更加符合人类的视觉特征.
A quantitative metric was proposed to objectively evaluate the quality of fused images.First, the Fuzzy C-Means clustering algorithm (FCM) was used to segment image in order to obtain the regional feature matrix.After dimensionality reduc- tion, we could calculate the distance between regions as regional similarity. Secondly, we used the ratio of the principal eigenvector derived from two original corresponding image regions as the weight of corresponding regions. The ratio of edge vectors amplitude between regions was used to denote local region saliency. Finally, we could attain the similarity between images. The results show that the proposed metric is more consistent with the human perception nature as it considers the local image variations and the regional saliency.
出处
《电子学报》
EI
CAS
CSCD
北大核心
2010年第5期1152-1155,共4页
Acta Electronica Sinica
基金
2006年教育部新世纪优秀人才计划(No.NCET-06-0487)
国家自然科学基金(No.60572034
No.60973094)
江苏省自然科学基金(No.BK2006081)
江南大学创新团队研究计划(No.JNIRT0702)
关键词
图像融合
融合评价
相似性
区域
image fusion
fusion evaluation
similarity
region