Abstract

This study, grounded in Waxman fusion method, introduces an algorithm for the fusion of visible and infrared images, tailored to a two-level lighting environment, inspired by the mathematical model of the visual receptive field of rattlesnakes and the two-mode cells' mechanism. The research presented here is segmented into three components. In the first segment, we design a preprocessing module to judge the ambient light intensity and divide the lighting environment into two levels: day and night. The second segment proposes two distinct network structures designed specifically for these daytime and nighttime images. For the daytime images, where visible light information is predominant, we feed the ON-VIS signal and the IR-enhanced visual signal into the central excitation and surrounding suppression regions of the ON-center receptive field in the B channel, respectively. Conversely, for nighttime images where infrared information takes precedence, the ON-IR signal and the Visual-enhanced IR signal are separately input into the central excitation and surrounding suppression regions of the ON-center receptive field in the B channel. The outcome is a pseudo-color fused image. The third segment employs five different no-reference image quality assessment methods to evaluate the quality of thirteen sets of pseudo-color images produced by fusing infrared and visible information. These images are then compared with those obtained by six other methods cited in the relevant reference. The empirical results indicate that this study's outcomes surpass the comparative results in terms of average gradient and spatial frequency. Only one or two sets of fused images underperformed in terms of standard deviation and entropy when compared to the control results. Four sets of fused images did not perform as well as the comparison in the Q^AB/F index. In conclusion, the fused images generated through the proposed method show superior performance in terms of scene detail, visual perception, and image sharpness when compared with their counterparts from other methods.

中文翻译：

---

昼夜两级光照环境下基于仿生响尾蛇视觉机制的图像融合

摘要

这项研究以Waxman融合方法为基础，引入了一种针对两级照明环境的可见光和红外图像融合算法，其灵感来自于响尾蛇视觉感受野的数学模型和两种模式细胞的机制。这里介绍的研究分为三个部分。在第一部分中，我们设计了一个预处理模块来判断环境光强度，并将照明环境分为白天和夜间两个级别。第二部分提出了专门为这些白天和夜间图像设计的两种不同的网络结构。对于可见光信息占主导地位的白天图像，我们将 ON-VIS 信号和红外增强视觉信号分别馈送到 B 通道中 ON 中心感受野的中心激励区域和周围抑制区域。相反，对于红外信息优先的夜间图像，ON-IR信号和视觉增强IR信号分别输入到B通道ON中心感受野的中心激励区域和周围抑制区域。结果是伪彩色融合图像。第三部分采用五种不同的无参考图像质量评估方法来评估融合红外和可见光信息产生的十三组伪彩图像的质量。然后将这些图像与相关参考文献中引用的其他六种方法获得的图像进行比较。实证结果表明，本研究的结果在平均梯度和空间频率方面均优于对比结果。与控制结果相比，只有一组或两组融合图像在标准差和熵方面表现不佳。四组融合图像在 Q ^AB/F指数方面的表现不如比较。总之，与其他方法生成的融合图像相比，通过该方法生成的融合图像在场景细节、视觉感知和图像清晰度方面表现出优越的性能。