A novel study on developing a model-driven visual dictionary