Visible and infrared image fusion (VIF) aims to fuse the complementary features of the source images into one image with richer information. The fused image has both contour and texture detail information from the visible image and the thermal radiation information of the target or background from the infrared image. Meanwhile, the color visible image has color information that is more in harmony with human's subjective visual perception. Most of the existing VIF methods focus on the processing of salient targets and texture details based on the grayscale channel, ignoring the fusion of the color information of the source images. This paper proposes a novel visible and infrared image fusion method Fusion with Preserved Hue, termed PHFuse, which can fuse and preserve abundant color information while fusing intensity and gradient information. Specifically, we design a new fusion process using the characteristics of the HSV color space, which performs more targeted fusion calculations on chromaticity and intensity information separately. In addition, we design a channel fusion branch, which can iterate the fusion strategy adaptively. We introduce the spectral angle mapping loss function in the fusion network, realize unsupervised training in fusing the color information. Subjective and objective evaluations on public datasets show that compared with current state-of-the-art methods, the PHFuse proposed in this paper can generate fused images with richer colors, and the intensity gradient information fusion performance is no less than that of the SOTA method.
Keywords: Color visible image; HSV color space; Image fusion; Infrared image; Preserved hue; Spectral angle mapping.
© 2025. The Author(s).