Scene Categorization by Introducing Contextual Information to the Visual Words