SDMatte is an interactive image matting method based on stable diffusion, which supports three types of visual prompts (points, boxes, and masks) for accurately extracting target objects from natural images.