In RGB-based hand pose estimation, occlusions caused by objects can degrade estimation accuracy and lead to tracking failures. This study proposes a real-time occlusion- resolution framework for hand pose estimation by integrating depth estimation, occluder region segmentation, and video inpainting. The core idea of the framework is to automatically identify occluders from input images and generate masks indicating regions to be inpainted. Experiments were conducted under tabletop occlusion scenarios by comparing inpainted and non-inpainted conditions. Performance was also evaluated using three hand pose estimation models. The results demonstrate that the proposed framework improves estimation accuracy regardless of the hand pose estimation model used and mitigates the adverse effects of occlusion.