Abstract: Conventional reverse time migration (RTM) suffers from amplitude distortion and insufficient resolution due to the approximation of the Hessian matrix as an identity matrix. Image-domain least-squares reverse time migration (LSRTM) with sparse regularization can effectively improve imaging resolution. However, existing sparse-regularized LSRTM methods generally neglect the spatial variability of seismic illumination. In poorly illuminated regions, weak but valid reflections are prone to being misidentified as noise and excessively suppressed, leading to the loss of deep and complex structural information. To address this issue, we propose an image-domain LSRTM method based on illumination-guided sparse regularization and total variation (TV) regularization. The proposed method utilizes seismic illumination energy to construct spatially varying sparse regularization parameters. Stronger regularization is imposed in well-illuminated regions to suppress noise and migration artifacts, whereas weaker regularization is applied in poorly illuminated regions to preserve weak reflection structures in deep and shadow zones. In addition, TV regularization is incorporated to enhance structural continuity and boundary preservation in complex geological settings. Tests on the Sigsbee2A model and field seismic data demonstrate that, compared with conventional image-domain LSRTM methods, the proposed method more effectively recovers structural details in poorly illuminated regions and improves both structural continuity and imaging resolution in complex geological environments, showing promising potential for practical applications.