Skip to Main Content
Although scene classification has been studied for decades, indoor scene recognition remains challenging due to its large view point variance and massive irregular artefacts. In fact, most existing methods for outdoor scene classification perform poorly in the indoor situation. To address the problem, we propose a hybrid image representation by combining the global information with the local structure of the scene. First, the global discriminative information is captured by pyramid GIST feature. Second, the local structure is encoded by the bag of features method with Histogram Intersection Kernel (HIK). Finally, HIK based SVM is employed for learning and classification. Experiments on the MIT indoor scene database show that our approach could significantly improve the recognition accuracy of the state-of-art methods by about 14%.