Saliency-aware Generative Art
Tao Wu · 2018
Generative art autonomously creates pleasuring images by using computer-based algorithms. In this paper, a novel framework of generative art is proposed by combining the techniques of saliency detection and stroke-based art. Given an input image as the reference target, the saliency map is evolved that will re-render the target image with non-photorealistic effects. An animation can be also generated by rendering the various strokes. The resulting animations have an interesting characteristic in which the target slowly emerges from a set of strokes. Various aesthetic features, including Benford's index, fractal dimension, global contrast factor and Shannon entropy are introduced into the evaluation framework of computational aesthetics. The results suggest that the proposed method can autonomously generate appealing images and animations with different styles by choosing different strokes, and it would inspire graphic designers who may be interested in subtle aesthetic patterns created automatically.