Real-time video matting using multichannel poisson equations
Minglun Gong, Liang Wang, Ruigang Yang, Yee‐Hong Yang · 2010
A This paper presennts a novel matti ing algorithm for processing video sequences in reall-time and onlin ne. The algorithm is based on a set of novel Poissson equations that are derivved for handlinng multichannel coolor vectors, as well as the depth informatioon captured. A simmple yet effectiv ve approach is also proposed to compute an inittial alpha matte in the color space. Real-timme processing speed is achieved thro ough optimizing the algorithm for parallel processinng on the GPUs. To process live video sequences online and autonomously, a mod dified backgrounnd cut algorithm is immplemented to separate foreground and backgroound, the result of which guides the automatic trimap generation. Quantitative evaluation on stiill images show ws that the alphaa mattes extracted using the presented algorithm is much more accurate than the onnes obtained using the global Poisson matting algorithm and are comparable to that of other state-of-the-art offline image mattinng techniques.