为了选取基于拉普拉斯金字塔变换融合方法效果较好的融合规则来提高多波段图像融合质量, 现结合 12种融合规则对可见光、红外短波和红外长波图像进行融合, 通过主观人眼视觉观察与客观评价中的信息量、统计特性和人眼视觉特性的评价指标对融合效果进行分析比较, 找到融合效果较好的融合规则。实验结果表明, 低频使用等值加权平均时, 融合效果较差, 对比度低。高频使用绝对值取大、低频使用区域匹配时, 融合结果清晰, 评价指标最高, 便于人眼观察。高频使用区域能量取大、低频使用区域匹配度取大时, 融合结果可以充分显示边缘和纹理信息。高频使用一致性检测和区域方差时融合效果一般。

A fusion rule based on the Laplacian pyramid transformation fusion method is required to improve multi-band image fusion quality. Twelve fusion rules were used to fuse the visible light, short-wave, and long-wave infrared images. Better fusion rules can be identified using comparison by subjective human visual observation and objective evaluation (such as amount of information and statistical properties).The experimental results show that using the low-frequency equivalent weighted average method, the fusion effect is poor and the contrast is low. Using the high-frequency absolute value maximization method and low-frequency regional matching methods, the fusion results are clear and the human observation evaluation index is the highest. The fusion results fully display the edge and texture information. The fusion effect is average when using high-frequency consistency detection and regional variance.