A visual analysis of the relationship between word concepts and geographical locations. CIVR, article 13

In this paper, we describe two methods to analyze the relationship between word concepts and geographical locations by using a large amount of geotagged images on the photo sharing Web sites such as Flickr. Firstly, we propose using both image region entropy and geo-location entropy to analyze relations between location and visual features, and in the experiment we found that concepts with low image entropy tends to have high geo-location entropy and vice versa. Secondly, we propose a novel method to select representative photographs for regions in the worldwide dimensions, which helps detect cultural differences over the world regarding word concepts with high geo-location entropy. In the proposed method, at first, we extracts the most relevant images by clustering and evaluation on the visual features. Then, based on geographic information of the images, representative regions are automatically detected. Finally, we select and generate a set of representative images for the rep-resentative regions by employing the Probabilistic Latent Semantic Analysis (PLSA) modelling. The results show the ability of our approach to mine regional representative photographs and cultural differences over the world