Robust multiband image segmentation method based on user clues

In this paper, a method for binary segmentation of multiband images based on a combination of dimensionality reduction techniques (Weighted PCA and Quadratic Programming Feature Selection), classification methods (Gaussian Mixtures Model and Random Forest), and a segmentation method (Quadratic Markov Measure Field Model) is presented. In this work, four pixels descriptors are addressed: Color, Discrete Cosine Transform (DCT), Gradient Fields (GF), and Adjacency Matrix (AM). Our method combines the outcome of several classifiers using an optimization criterion which results in a robust method for image segmentation based on user clues. We evaluate our method capabilities with different image types such as RGB and satellite images. Experimental results demonstrate our method's performance.