多传感器图像融合的客观评价与分析

图像融合作为一种有效的信息融合技术，已经广泛用于军事、遥感、机器视觉和医疗诊断等领域。这里研究了多传感器图像融合效果的客观评价方法，其中包括均方根误差、平均误差、灰度标准差、熵、熵差、交叉熵、互信息和空间频率。将加权平均、金字塔、小波变换三种方法用于多传感器图像的融合，实验结果表明客观评价结果与视觉评价的结果是一致的，是一种行之有效的评价手段。

Objective Evaluation and Analysis of Multisensor Image Fusion

Being an efficient method of information fusion, image fusion has been used in many fields such as machine vision, medical diagnosis, military applications and remote sensing. The objective evaluation of fused image is studied. The objective evaluation measures are root of mean square error, mean error, standard deviation, entropy, entropy difference, cross entropy, mutual information and spatial frequency. Three image fusion algorithms, weighted mean, Laplacian pyramid based, wavelet decomposition based, are applied to fuse several images. Experimental results show that conclusions of objective evaluation coincide with those of human vision evaluation. So the objective evaluation measures are feasible and efficient.