A real-time professional photographing guiding system through image composition analysis

Abstract

In professional photographing, a valuable photo usually comprises of a good image composition because people prefer well-composed photos. In the field of image aesthetics, image composition plays an important rule for predicting whether an image is taken by professional or not. As a result, to help amateurs take good photographs, a real-time system that analyzes the image composition in the field of view of a camera in real-time is helpful. In this paper, we proposed a real-time image composition analysis system to help guide users take better photos in real time by showing the analyzed results on screen constantly and instantly. Several image compositions, which are central, rule of thirds, horizontal, vertical, diagonal and perspective, can be classified instantly with our proposed methods. We use the saliency map and prominent lines as the main features for analysis, and use multilayered perceptron to predict the image composition of an input image. The accuracy is above 97%, which is higher than most of the state-of-the-art methods.