A Context-Sensitive Technique Based on Support Vector Machines for Image Classification
In this paper, a novel context-sensitive classification technique based on Support Vector Machines (CS-SVM) is proposed. This technique aims at exploiting the promising SVM method for classification of 2-D (or n-D) scenes by considering the spatial-context information of the pixel to be analyzed. The context-based architecture is defined by properly integrating SVMs with a Markov Random Field (MRF) approach. In the design of the resulting system, two main issues have been addressed: i) estimation of the observation term statistic (class-conditional densities) with a proper multiclass SVM architecture; ii) integration of the SVM approach in the framework of MRFs for modeling the prior model of images. Thanks to the effectiveness of the SVM machine learning strategy and to the capability of MRFs to properly model the spatial-contextual information of the scene, the resulting context-sensitive image classification procedure generates regularized classification maps characterized by a high accuracy. Experimental results obtained on Synthetic Aperture Radar (SAR) remote sensing images confirm the effectiveness of the proposed approach.
KeywordsSupport Vector Machine Image Classification Synthetic Aperture Radar Markov Random Field Synthetic Aperture Radar Image
- 3.Platt, J.C.: Probabilistic Outputs for Support Vector Machines and Comparison to Regularized Likelihood Methods. In: Smola, A., Bartlett, P., Scholkopf, B., Schuurmans, D. (eds.) Advances in Large Margin Classifiers. MIT Press, Cambridge (1999)Google Scholar