Least square image matting

This paper addresses the well-known problem of natural image matting. Most of the previous matting algorithms require the user to define the tri-map, which is an inconvenient work and sometimes a burden, especially in a complex situation. This paper uses ceratain user defined foreground and background strokes to estimate the image matte. First we use a Gauss Markov random field to model the matting problem. Then we use the least square optimization approach to solve it. Experimental results show that our approach could properly handle confused boundaries. It also could deal with semi-transparent conditions such as fire etc.