使用多尺度视觉注意提取的双波段图像融合

Dual-Band Image Fusion Using Visual Attention Extraction with Multiple Windows

针对可见-红外图像,结合多窗口选择下的视觉注意图的提取,实现了双波段图像的增强融合。利用局部频率调谐获取视觉注意图,以模拟人眼赋予不同区域与像素不同的注意权重的功能;设计了多窗口尺度下的视觉注意图的提取方法,以凸显提取不同尺寸的特征信息;在多窗口尺度下的视觉注意图的基础上实现双波段图像的增强融合。利用双波段图像对进行论证实验,并结合主客观评价对多种融合方法进行评估。实验结果表明,所提出的方法在满足人眼的视觉要求并有效保留细节的同时,具有较快的运算速度。

For the visible and infrared images, dual-band image fusion enhancement is achieved with visual attention extraction under multiple windows. The visual attention map is obtained using local frequency tuned idea, and this attention map simulates human visual system to give different visual weights to different areas and pixels. The method for multiple window-based visual attention maps extraction is designed, which could highlight feature information with different sizes. Based on multi-scale visual attention maps, dual-band image fusion enhancement is achieved.Using dual-band image of demonstration experiments, and combining with both subjective and objective assessment to evaluate various fusion methods. The experimental results demonstrate that the proposed method can efficiently reserving details and meet the need for human visual at the same time, and has faster operating rate.