为了在多波段图像融合中选用合适的直觉模糊化方法来处理不确定性问题, 对 5种常用的直觉模糊集方法在多波段图像融合中的应用进行了比较。首先将多波段图像进行直觉模糊化处理, 对隶属度图像进行去模糊化得到直觉模糊图像; 然后, 将直觉模糊图像进行非下采样轮廓波变换(Nonsubsampled contourlet transform, NSCT), 对低频图像进行直觉模糊化处理, 所得隶属度作为权重进行加权融合, 高频使用取大规则进行融合; 最后通过逆变换得到融合图像。对融合结果采用主观人眼视觉观察和客观评价指标体系进行分析比较, 得到较好的直觉模糊集方法的优势性能。实验结果表明, 与 Sugeno、Yogita、Yager及 Chaira四种直觉模糊化方法相比, Bala直觉模糊化方法可以有效提高融合结果的亮度和对比度, 而且融合结果边缘清晰, 纹理特征明显, 具有更好的视觉融合效果和客观质量评价。

To determine the right intuitionistic fuzzy method in multiband image fusion, which can deal with the uncertainty problem, in this paper, five commonly used methods of intuitionistic fuzzy sets and their applications are compared. First, the multiband images were processed by the intuitionistic fuzzy sets, and the images of membership degree were blurred to obtain the intuitionistic fuzzy images. Then, the nonsubsampled contourlet transform (NSCT) was performed on intuitionistic fuzzy images, the low-frequency images were processed by the intuitionistic fuzzy sets and fused by the weighted average, where the weight is the membership degree, and the high-frequency images were fused by take-the-biggest-of-all. Finally, the fused image was constructed by the inverse NSCT. Both subjective and objective evaluations were compared to determine the best-performing intuitionistic fuzzy set method. The experimental results show that compared with Sugeno, Yogita, Yager, and Chaira methods, the Bala intuitionistic fuzzy method effectively improves the brightness and contrast of the fusion results. The fusion results have a clear edge and the texture feature is apparent, which offer a better visual effect and objective evaluation.