This paper systematically introduces the fundamental principles of traditional and single-photon LiDAR systems, and image fusion technologies. It provides an in-depth analysis of the feature disparities between traditional / single-photon LiDAR point clouds and visible images, as well as the associated image registration challenges. The study discusses the advantages and limitations of current registration methods, including calibration-based projection, feature matching, and deep learning approaches. Furthermore, it reviews the state-of-the-art in fusion technology for these modalities and explores its applications in fields such as target detection, recognition, and three-dimensional reconstruction. Looking ahead, the development of LiDAR point cloud and visible image registration necessitates research into methods capable of automatic adjustment and calibration during practical deployment. The application of more sophisticated neural network models is required to achieve deep multimodal information fusion, alongside further algorithm optimization to enable large-scale implementation. Regarding fusion technology, future work should focus on front-end collaborative design at the hardware level, adopt more advanced active sensing paradigms to enhance system efficiency and intelligence, and leverage advanced deep learning-based super-resolution networks to improve perceptual capabilities in complex environments.