Latent visual context learning for web image applications

Recently, image representation based on bag-of-visual-words (BoW) model has been popularly applied in image and vision domains. In BoW, a visual codebook of visual words is defined, usually by clustering local features, to represent any novel image with the occurrence of its contained visual words.