Seamless Capturing of Moments Using Photographic Compositions and Image Aesthetics

摘要

This paper proposes an automatic photograph system, which aims to take good photos. We suppose that the good photos are aesthetically pleasing and visually balanced. In fact, the term, visually balanced, have been defined to photographic composition rules. In order to achieve this, photographic composition classification, image aesthetic enhancement, and aesthetic assessment using deep learning models are introduced in this paper. The proposed method classifies nine photographic compositions for captured images, and enhances aesthetic quality of the images classified to one of nine compositions. Then, aesthetic scores are measured for the enhanced images to determine that the image is able to be a good photo. The proposed method can be widely applied to all smart devices possessing cameras, and thus is able to automatically collect memorable moments of daily life.