Content and context features for scene image representation

Existing research in scene image classification has focused on either content features (e.g., visual information) or context annotations). As they capture different information about images which can be complementary and useful to discriminate of classes, we suppose the fusion them will improve results. In this paper, propose new techniques compute features, then fuse together. For design multi-scale deep based background foreground images. use annotations similar available web a filter words (codebook). Our experiments three widely used benchmark datasets using support vector machine classifier reveal that our proposed produce better results than existing respectively. The two types significantly outperform numerous state-of-the-art features.