Analytical Aesthetics Applied to Large Collections of Works of Visual Arts

Experiments show that design elements used by artists are associated with mental states. We aim to utilize this relationship in both design elements trigger certain mental states, and mental states are expressed by suitable design elements. To describe and utilize this relation, we propose a system of signs defined by sets of images. For each set, each of the images defining it evokes the same associations in viewers. The creation of this system comprises the following
Definition of a sign as a set of images
Definition of the meaning of the signs
Combination of signs
Relationships between the signs
To solve these tasks, we calculate the similarity of images in terms of human perception using the algorithm given in [1]. This algorithm is designed to create sequences of fixations that mimic human eye movement. Based on these fixations, it calculates image similarity. Following the definition of the sign system, we describe an analytic aesthetics (cf. [2]). The function of our analytic aesthetics maps every image to a unique point in a feature space. Each sign from the above-mentioned sign system represents one dimension in this space. The value assigned each dimension is the measure of similarity between the image and the respective sign. To show the effectiveness of the proposed technique, we have implemented a working software system that includes a visual aesthetic based on a system of signs built on a corpus of 3.500 pictures of visual arts. Using this system, we demonstrate how our technique facilitates the search, exploration, and general structuring of our corpus.
[1] Hermann Pflüger, Benjamin Höferlin, Michael Raschke, Thomas Ertl. Simulation Algorithm for Fixations during Perception of Visual Arts. ACM Trans. Appl. Percept., in preparation.
[2] Frieder Nake. 1974. Ästhetik als Informationsverarbeitung. Springer Verlag Wien New York.