Abstract: Traditional reverse time migration (RTM) is essentially the adjoint operation of the wave-equation forward modeling operator, and its imaging results are affected by the Hessian matrix, resulting in amplitude distortion and limited resolution. Image-domain least-squares reverse time migration (LSRTM) with sparse regularization can effectively improve image resolution; however, under complex geological conditions, conventional uniform sparse regularization tends to suppress weak but valid reflections in poorly illuminated regions, leading to structural information loss. To address this issue, we propose an image-domain LSRTM method based on illumination-guided sparse regularization and total variation (TV) regularization. First, seismic illumination energy is used to construct spatially adaptive sparse regularization parameters. Stronger sparse constraints are applied in well-illuminated regions, while weaker constraints are assigned to poorly illuminated regions to preserve valid weak reflections in deep and shadow zones. Second, TV regularization is introduced to improve structural continuity and preserve boundaries in complex geological structures. Finally, the optimization problem is solved using the alternating direction method of multipliers (ADMM) for efficient inversion. Tests on the Sigsbee2A model and field data demonstrate that, compared with conventional image-domain LSRTM, the proposed method effectively improves structural continuity and imaging resolution in weakly illuminated regions, showing strong potential for application in complex geological settings.