Analyzing art works: the six steps methodology

Abstract

Art analysis is a key aspect for computational systems whose goal is to generate visual artifacts. This paper proposes a six steps methodology to analyze and represent design principles from art works. Our approach starts with an image segmentation followed by the construction of a straight skeleton. Then we extract some color information and perform shape analysis and classification. Finally some design principles are calculated and groups of elements are built over a complete graph. Our internal art work representation gives us a way to approximate the phases of artistic appreciation proposed by some authors. We show a procedure to generalize compositional rules for the generation of new abstract art works based on the steps of the proposed methodology. We plan to use a self organizing map to cluster our art work representations and use this information to build a hypergraph and/or multigraph. Since these graphs can represent design principles, the system will be able to use these structures to explore new ways to generate artifacts and measure their novelty compared to previous exemplars.