Numerical Analysis and Comparison of the Algorithms of Image Foreground Extraction

Abstract

Foreground extraction is widely used in pattern recognition, movie and arts, etc. And the most important part of Foreground extraction is detecting the image edge. By detecting the image edge, we can build the models to extract foreground from the image through region, pixel or color. This paper begin with the analysis and evaluation of several typical algorithms of edge detection operator, and then the mixture Gaussian model which based on the foreground and background colors basic model is given to balance the insufficient of two models. Next we get matrix E(x) from the mixture Gaussian model, and then realize foreground extraction by maximum flow minimum cut algorithm cutting. Finally as the parameter's change, the extracted effect is analyzed.