This paper proposes a novel, depth-based image refocusing system for mobile devices equipped with stereo cameras, called Scribble2focus. The system provides a simple and intuitive interface, which allows the user to refocus the image by indicating an object of interest with a simple finger stroke. A stereo matching algorithm based on a Minimum Spanning Tree approach is proposed to infer depth from images from the stereo cameras. Using the depth information and a thin lens model, input all-focus images are synthetically refocused. Experiments show that the proposed system produces quality refocusing results with minimal computational requirements. The system requires only around 1.5 seconds to process a pair of 640×360 images in the off-line step, making it possible for future deployment on a mobile device.