A Complexity Approach for Identifying Aesthetic Composite Landscapes

Abstract

The present paper describes a series of features related to complexity which may allow to estimate the complexity of an image as a whole, of all the elements integrating it and of those which are its focus of attention. Using a neural network to create a classifier based on those features an accuracy over 85% in an aesthetic composition binary classification task is achieved. The obtained network seems to be useful for the purpose of assessing the Aesthetic Composition of landscapes. It could be used as part of a media device for facilitating the creation of images or videos with a more professional aesthetic composition.

Photo Composition with Real-Time Rating

“…They later on proposed to refine the photo ranking system for amateurs by an interactive personalization approach [4] so that each person gets to define his or her own ranking through positive and negative feedbacks. Carballal et al [5], with the target being landscape images, proposed some features and metrics that could be used for estimating the complexity of an image and a binary classifier is trained through the help of an expert to judge an image's aesthetic quality with the the accuracy of 85%. Obrador et al [6] proposed some low-level image features such as the simplicity of a scene, visual balance, golden mean and golden triangles and managed to achieve comparable aesthetic judgement, compared with the start-of-the art of their times.…”