Probabilistic Models for Supervised Dictionary Learning

Dictionary generation is a core technique of the bag-of- visual-words (BOV) models when applied to image cate- gorization. Most of previous approaches generate dictio- naries by unsupervised clustering techniques, e.g. k-means. However, the features obtained by such kind of dictionaries may not be optimal for image classification. In this paper, we propose a probabilistic model for supervised dictionary learning (SDLM) which seamlessly combines an unsuper- vised model (a Gaussian Mixture Model) and a supervised model (a logistic regression model) in a probabilistic frame- work. In the model, image category information directly affects the generation of a dictionary. A dictionary ob- tained by this approach is a trade-off between minimization of distortions of clusters and maximization of discriminative power of image-wise representations, i.e. histogram repre- sentations of images. We further extend the model to incor- porate spatial information during the dictionary learning process in a spatial pyramid matching like manner. We ex- tensively evaluated the two models on various benchmark dataset and obtained promising results.