由于基于经典Waxman融合模型得到的伪彩色融合图像目标不够清晰, 本文提出了一种改进的感受野融合模型。对红外图像和可见光图像分别进行了ON对抗增强和OFF对抗增强; 将红外ON对抗增强图像馈入中心-环绕感受野模型的中心兴奋区域, 可见光OFF对抗增强图像馈入环绕抑制区, 得到融合图像的B分量; 将红外OFF对抗增强图像馈入中心-环绕感受野模型的环绕抑制区域, 可见光ON对抗增强图像馈入中心兴奋区, 得到融合图像的G分量; 将可见光ON对抗增强图像直接作为融合图像的R分量; 然后, 输出RGB伪彩色融合图像。最后, 用Waxman方法和本文提出的方法分别对两组源图像进行融合, 并用信息熵和平均梯度对融合结果进行了评价。结果表明, 采用提出的模型, 第一组融合图像的信息熵和平均梯度比Waxman融合模型分别高出0.314 6和0.004 1, 第二组融合图像的信息熵和平均梯度比Waxman融合模型分别高出0.255 1和0.002 7。得到的数据显示本文提出的融合模型的融合效果优于经典Waxman模型。

Since the target in a pseudo color fusion image based on the classical Waxman fusion model is not clear, this paper proposes an improved receptive field fusion model. The infrared image and visible light image were respectively ON against enhanced and OFF against enhanced. The infrared ON against enhanced image was fed into an center exciting area of the center-surround receptive field model and the visible light OFF against enhanced image was fed into a surround inhibition zone to get the fusion image B component. Then, the infrared OFF against image was fed into a center inhibition zone of the center-surrounding receptive field model and the visible light ON against enhanced image was fed into the center exciting area to get the fusion image of G component. Furthermore, the visible ON against enhanced image was directly taken as the R component of fused image and to output the pseudo color fusion image. Finally, the fusion experiments were performed for two groups of original images by Waxman model and proposed model and the fusion results were evaluated by the information entropy of fused image and the average gradient. The results show that the first set data by proposed method are higher 0.314 6 and 0.004 1 respectively than that of Waxman fusion model, and that of the second set data by proposed method are higher 0.255 1 and 0.255 1 than that of the Waxman fusion model. It concludes that fusion effect of the proposed fusion model is superior to that of the classical Waxman model.