Airborne laser scanning (ALS) is a tool now widely used in archaeology [1–8], geomorphology, and earth sciences [9–11] to detect natural landforms or remains of human activity, especially in forested areas, where other remote sensing techniques are unsuccessful or time-consuming. The main interest of this technology is to cover large areas while offering high spatial resolution capabilities. Research programs using LiDAR data are becoming more and more frequent. These studies are very often based on a multidisciplinary approach, involving specialists in archaeology, forestry, geomorphology, volcanology [12]. After ALS data acquisition, a point cloud classification has to be carried out, and the resulting Digital Terrain Model (DTM) and Digital Surface Model (DSM) areas are produced. Different visualization techniques are then generally applied to the DTM, to enhance micro-topography versus global topography and help to the detection of target features. The most common are multidirectional oblique weighting hillshade (MDOW), slope [15], Local Relief Model (LRM) [16,17], Sky-View Factor (SVF) [18], positive and negative openness [19,20]. These methods can be divided into two main categories: hillshade, Sky-View Factor, and openness are typically illumination techniques, based, respectively, on the sky portion visible from each position or on the openness characteristics of the relief at each position. They allow highlighting high-frequency components of the relief but remove all the elevation information. Indeed, these methods do not directly restore the topographic variations but rather the consequences of these variations, such as the portion of the visible sky. On the contrary, slope and LRM are DEM manipulating methods. They consist in computing elevation characteristics parameters, respectively, the slope or the high-frequency component of the relief. Recent studies proved that LRM is one of the most efficient visualization techniques [21,22]. The basic principle is to apply moving average filtering to the DTM, in order to remove the general trend of the natural relief: the local relief, characterized by sharp variations, is then revealed [16]. The extent of this filtering (kernel- or window-size) has to be defined by the user, according to both global relief characteristics and morphometric characteristics of the target features. The choice of the correct filtering extent is important but not critical when it is applied to the detection of well-preserved anthropogenic remains. This is because they are generally characterized by sharp changes in local relief, corresponding to the high-frequency component in the frequency domain, well separated from the lower frequency component (features of the natural relief). However, when the aim is to detect all the potentially interesting features, including geomorphological shapes or eroded anthropogenic remains, the filtering perimeter has to be adapted to the characteristics of the natural relief (e.g., slope), which influence the performance of the LRM significantly. Indeed, it is only possible to detect an artifact if it provides a sufficient contrast compared to the surrounding features, i.e., if its frequency signature is significantly higher than the one of the natural reliefs [23]. As LiDAR detection is now used on very large areas, several LRM configurations need typically to be used in order to detect both slight and sharp local relief variations in complex topography contexts, including flat areas and medium to steep slope areas, after what the results from the different models could be eventually merged. This process can be confusing and time-consuming, especially for inexperienced users, and also introduces significant bias, as the decision of the configurations to be tested depends on the skills (and the available time) of the operator. The Self-AdaptIve LOcal Relief Enhancer (SAILORE) approach present an evolution of the widely used Local Relief Model method, allowing the automatic adaptation of the filtering size according to natural relief, producing a single-model, which makes simpler, faster, more efficient, and more reliable detection of target features in large datasets with variegated topography. It automatically uses the best filter configuration, allowing the detection of all the types of anthropogenic remains, independently of the global relief context.