Improved algorithm on natural image matting

The natural image such as hair and smog in one image is often thinner than a pixel, which makes lots of pixels, located at the boundary region between the foreground and the background in one discrete digital image, a composition of the foreground color and the background color. As to the segmentation of these images, the goals of matting can't be achieved with normal methods. Natural image matting based on alpha channel is targeted to such issue, which can separate the foreground and the background of a single pixel according to its color. So the effect of this method is normally better than other that of traditional methods. A new natural image algorithm was put forward based on alpha channel with a good result as well as a rapid speed. Experiment results demonstrate that this approach has the advantages of fast execution, good result, as well as easy to use.