A new contextual version of Support Vector Machine based on hyperplane translation

Support Vector Machine (SVM) is a method widely used for image classification. The original formulation of this method does not incorporate contextual information. This study brings a new perspective regarding contextual SVM. The main idea of the presented proposal consists on translates, individually for each pixel using it contextual information, the separation hyperplane originally designed by SVM. A case study using ALOS PALSAR image shows that the proposed method produces better results than traditional SVM.