MAGMA—efficient method for image annotation in low dimensional feature space based on Multivariate Gaussian Models

Automatic image annotation is crucial for keyword-based image retrieval. There is a trend focusing on utilization of machine learning techniques, which learn statistical models from annotated images and apply them to generate annotations for unseen images. In this paper we propose MAGMA - new image auto-annotation method based on building simple multivariate Gaussian models for images. All steps of the method are thoroughly described. We argue that MAGMA is efficient way of automatic image annotation, which performs best in low dimensional feature space. We compare proposed method with state-of-the art method called continuous relevance model on two image databases. We show that in most of the experiments simple parametric modeling of probability density function used in MAGMA significantly outperforms reference method.