Computational Aesthetic Based on Fuzzy logic and Neural network for Aesthetic Quality Assessment of Photographic Image **

Abstract - Humans have the intelligence and ability to assess and classify images based on visual perception in a certain aesthetic categories. How to model the human capabilities into computer systems become major challenges in aesthetic computing, a new field study in computer science. In photographic field, there are some standard parameters that are used by professional photographers to classify images based on aesthetic parameters; lighting, clarity of the contrast, composition, and simplicity. Based on these parameters, this research focuses on developing intelligent systems for classifying photographic images by using ANFIS (Adaptive Neuro-Fuzzy Inference System) method. Systems based on fuzzy conclusion that utilize fuzzy if-then rules can model the qualitative aspects of human knowledge and reasoning processes without the benefit of giving a quantitative analysis of the right, so that the ANFIS method has possibility to capture human aesthetic knowledge on image assessment. Research method begin with subject extraction for image, then continue with feature extraction based on four aesthetic parameters to get training and testing data for ANFIS training and testing. Based on experimental results, image classification system using aesthetic parameters based on fuzzy inference system and neural network through the ANFIS method allows to be developed further. Test results based on empirical test results show the system accuracy of 75% and 25% error rate.

Index Terms— aesthetics, image, photography, fuzzy logic, neural network.

** This research of thesis has been published in SITIA (Seminar of Intelligent Technology and Its Application) 2010 in Surabaya, and ICCI (International Conference on Creative Industry) 2011 in Bali.