Multi-Class Labeling Improved by Random Forest for Automatic Image Annotation

Recently automatic image annotation (AIA) has been arising as a key technology to support image retrieval. The representative algorithm is Semantic Multiclass Labeling (SML [1]), which constructs a parametric generative model of a distribution of local image features in a class with a gaussian mixture model. Although SML shows good accuracy, SML has not been used widely because of its long training time and annotation time. In this paper we propose a method of improving SML by dealing with Random Forest instead of the gaussian mixture model. We evaluate our proposal by using the standard corpus, Corel5K and IAPRTC-12. The experimental results demonstrate that our method can train very fast and annotate multiple labels very fast, with keeping comparable performance as existing methods.