Automatic image creation via artistic composition principles

Methods for choosing image parameters in both art and computer graphics are currently subjective. The choice of parameters results in images of varying quality. One aspect of image quality is the composition of the image. While the principles underlying composition are somewhat subjective, a portion of the compositional rules can be approximated quantitatively. We use this quantification to design an objective function and use numerical optimization to automatically arrive at images with acceptable composition. For a given subject or scene, the optimization procedure chooses format, view point, layout, and lighting parameters. The resulting image is determined by characteristics the objective function rewards. We show several images generated using such optimization, and argue that these images have good composition.