Optimal viewpoint selection based on aesthetic composition evaluation using Kullback-Leibler divergence

Abstract

In this paper, we construct a robot photographic system to search for the optimal viewpoint of a scene. Based on some known composition rules in the field of photography, we propose a novel aesthetic composition evaluation method by the use of Kullback-Leilber divergence. For viewpoint selection, we put forward a method called Compositionmap, which can estimate the aesthetic value of scenes for each candidate viewpoint around the target group. At last, the effectiveness of our robot photographic system is confirmed with practical experiments.