Statistical-Based Image Tagging

The automation of image tagging is extremely important research topic in recent years due to its importance in building large image databases. The optimal goal of recent research is to automatically annotate images and overcome the semantic gap between the image content and the associated text representation. Image retrieval from large databases is one of the important domains that can benefit from automatic tagging. The automatic tagging task is currently associated with many challenges ranging from inaccuracy of retrieval technique to the efficiency and speed of the tagging approaches. In this paper we propose a statistical based tagging approach that uses normalized multidimensional color histograms as a global descriptor of low level features of images. Our results demonstrate that our proposed approach can outperform the Learning based methods in terms of accuracy and speed.