An intelligent clustering method for highly similar digital photos using pyramid matching with human perceptual 25 color histogram

Abstract

Recently, as the number of photos to be managed grows, photo classification becomes one of the most burdensome tasks. Besides, these technical advances encourage people to take duplicate photos for the more clear and the more user-wanted photos. This paper presents an automated clustering method to classify hundreds of photos considering the people's recent photographing behavior. First, we partition the input photo sets into trivial event groups. Then, we employ an interval graph considering their color similarity from temporally consecutive photos to construct each similar photo group. For this clustering, we used 25 color block histogram based on pyramid matching. The user experiment shows that our algorithm is enough correct to classify hundreds of photos. © 2011 Springer-Verlag.