Research on a Real-Time Automatic Image Matting
Wei Zhou · Journal of Southwest China Normal University · 2011
Image matting is a fundamental problem in image processing and is,in essence,an ill-posed problem.Most state-of-the-art algorithms require user interactions(such as the supply of trimaps or scribbles) to generate high-quality output.The authors of this paper propose a real-time automatic image matting method,which uses the input left and right images to generate a depth map through real-time stereo vision.Then,based on the depth map,a rough trimap is obtained by threshold segmentation.Finally,this rough trimap is used to guide the closed-form solution image matting and achieve more accurate results.Owing to its high-speed and high-quality,this method is desirable for real-time applications.