Proposal of Situation-based Clustering of Sightseeing Spot Images based on ROI-based Color Feature Extraction

Abstract

As the amount of photos shared on album websites grows up rapidly, the meaningful grouping of these images becomes important. In particular, sightseeing spot images in different situations, such as weather conditions and seasons, are useful for tourists to decide when to visit there. This paper proposes a method to group sightseeing spot images into several different situations by a hierarchical processing based on color feature on the designed region (ROI, region of interest) and K-means clustering. At the first stage, night view images are discriminated from daytime images. Then the daytime images are divided into sunrise/sunset and other images at second stage. Finally, the cloudy images are separated from shine images in others images generated at second stage. Experimental results show that the extraction of color feature within ROI is effective for obtaining clusters with high precision and recall.