Computational research on flat image design and style based on perceptual feature quantification

Abstract In order to solve the problem that a large number of low-value and easy-to-consume graphic designs consume human financial resources society also waste inherent value design resources, by dividing layers pixels, elements, relations, planes applications, this article defines corresponding feature representations studies related methods for quantifying geometric features, perceptual features style features. The content includes extraction method element colour, calculation layout-aware colour-match-aware feature, pairwise comparison feature. experimental results show proposed in has lower complexity is significantly better than other algorithms. Comparing with methods, average time algorithm 2.6 s under condition guaranteed images. study feature-driven image not only great significance solving repeated problems but provides basis reference future development intelligent design.