Significance Information is the basic element of human cognition of the world. Notably, 83% of human access to information relies on vision; hence, imaging technology based on image information acquisition is important for national defense and improving people ' s lives. Traditional imaging processes entail a geometric-optical transformation relationship from the target scene to the imaging sensor, facilitating accurate recording of object information. As the imaging technology has evolved, diverse imaging equipment has overcome the limitations of human vision, enabling us to perceive finer, more distant, and broader phenomena. For instance, microscopes allow us to observe the mysterious microworld and astronomical telescopes grant us a glimpse into the expansive cosmos spanning billions of light-years. Moreover, infrared thermal imaging cameras overcome the darkness by enhancing our vision in low-light conditions. However, regardless of the advancement of imaging equipment, when a scattering medium exists between the target scene and imaging apparatus, issues arise due to scattering and absorption. On the one hand, normal imaging light gets blocked, leading to signal energy loss. On the other hand, some of the signal light from the target scene gets disturbed, causing deviation in the original geometric imaging model of point-to-point imaging. In addition, scattered ambient light (referred to as airlight) enters the imaging detector. This light is not encoded by the target scene, does not carry any target information, and is additive noise. Research on scattering imaging has been conducted since the 1950s, with China joining the international efforts around 2000. Since then, advancements in hardware and algorithm development have led to the application of numerous new technologies. However, improving detection depth, expanding the field of view, increasing the imaging speed, and improving the recovery quality remain crucial scientific challenges. Scattering imaging is a crucial technique for image acquisition in complex scenes. Over the past decades, many methods have been proposed to address this practical problem. These methods can generally be categorized as active or passive based on whether or not active illumination is required. Among active methods, the most straightforward and useful method is to select the light that has been scattered the least by gating, wavefront compensation, and pointwise scanning and use it in florescence microscopy. These methods have broad applications across various fields. However, the imaging distance/depth achievable by these methods in scattering media is limited due to the attenuation of ballistic light. To further improve the imaging depth, active methods such as optical phase conjugation, wavefront shaping, optical transmission matrix measurement, speckle correlations based on optical memory effects, and deep learning have been proposed to exploit scattered light for image formation. Passive methods do not rely on active illumination. In particular, scattering particles absorb and scatter light from the object of interest and generate a substantial amount of airlight by directly scattering light from the illumination source, such as the sun. The presence of airlight considerably reduces the contrast of captured images, resulting in poor visibility. Conventionally, image dehazing algorithms are applied to enhance contrast. These algorithms can be broadly categorized into two groups. The first group includes image restoration algorithms based on physical models, such as polarization models, image depth priors, and dark channel priors. The second group comprises image enhancement algorithms that do not depend on physical principles; typical examples include Retinex-based algorithms, wavelet transform, and data-driven deep learning. Progress The research progress of various scattering imaging techniques is summarized in Table 1. These techniques are classified into two imaging modes, namely active and passive, based on whether active light illumination is required or not. This classification stems from the different noise models that contribute to image degradation. In the active mode, most scattered light carries disturbed object information or scattering channel information, representing invertible multiplicative noise. Consequently, it can be used to extract ballistic light or utilize scattered light for imaging. Conversely, in the passive mode, scattered light primarily originates from airlight randomly scattered by the atmosphere, constituting additive noise without object signals. Therefore, most current research focuses on dehazing algorithms for extracting ballistic light. Based on different modes and utilization of scattered light for imaging, the summary proceeds as follows: first, techniques for extracting ballistic light in the active mode (Fig. 1) are summarized as simple and effective methods with a high degree of required engineering (Figs. 2 and 5). Next, the summary covers imaging techniques using scattered light in the active mode and non-line-of-sight imaging (Fig. 10). Subsequently, the summary addresses imaging techniques used for extracting ballistic light in the passive mode (Fig. 13). Finally, examples of relevant applications of deep learning in both the modes are presented (Figs. 18 and 19). Conclusions and Prospects Most of the scattered light in the active mode constitutes invertible multiplicative noise, enabling its utilization for extracting ballistic light or achieving imaging using scattered light. Conversely, scattered light in the passive mode primarily consists of additive noise, necessitating reliance on dehazing algorithms for extracting ballistic light. The application of deep learning as a new technology has demonstrated numerous examples of its effectiveness in both the modes. Looking ahead, advancements in optoelectronic devices and algorithmic arithmetic are anticipated to enable improved integration of the front-end and back-end in scattering imaging research. This integration is expected to generate innovative ideas, leading to the further improvement of imaging depth and recovery quality.
