Soil structure, whatever the scale, is a key element to understand soils functioning. The quantitative characterization of the pore network and its dynamics, together with soil organism activity and interactions with their environment, provides an understanding of the links between soil structure and functions (Konig et al., 2016, Porre et al. 2016). Cutting-edge imaging methods are thus increasingly needed and developed. They are diverse, from non-invasive (X-ray or neutron (micro)tomography) to destructive (optical, electronic, chemical microscopy…), in constant evolution, increasingly accessible, and produce always more data. It can be 2D or 3D images, time series images (to follow dynamic processes) or one-time images (to study the soil at a given time). To centralize these images in a unique information system, to facilitate and secure their storage along with their metadata, their publication and citation, and promote their sharing with the scientific community, we are building the ImageSoil database, which aims to be complementary to the Soil Structure Library (Weller et al., 2021. ImageSoil stores raw or processed images with i) their pedological metadata (sampling location and date, soil profile/horizon description…), ii) image acquisition metadata (image tools including beam characterization, pixel/voxel properties...), and iii) image analysis metadata (software, segmentation thresholds....). Soil images can also been linked through their metadata to National Information System storing soil properties. With ImageSoil, we aim to enable soils scientists worldwide to share their soil image data and promote further analyses of these newly available images, especially (but not exclusively) to answer questions requiring soil structure information.