Research On Visual Perception Intelligence Based Graphic Design

As a common type of visual communication,graphic designs,such as posters,ad-vertisements,magazine covers,infographics,are widely used in life.Designers use the combination of images,texts and symbols to display information harmoniously.The aes-thetics of graphic designs is closely related to various design elements,such as layout,background image,color,font.The factors such as the arrangement,combination and interaction bring readers different feelings and aesthetic experience of obtaining infor-mation.Considering some actual scenarios in graphic design area,this paper studies the importance of visual perception and aesthetic rules in graphic design methods.Automatic typography is a practical but challenging problem.It helps designers avoid highly repetitive tasks and amateur users achieve high-quality textual layout designs.In this paper,we propose an efficient deep aesthetics learning approach to generate harmo-nious textual layout over natural images.Our method incorporates both semantic features and visual perception principles.We propose a semantic visual saliency detection network combined with a text region proposal algorithm to generate candidate text anchors with various positions and sizes.We develop a new text region proposal algorithm based on the text-driven probability map instead of the original image.After generating text region proposals,we need to estimate the aesthetic score of each result.We build a deep scoring network with a multi-scale feature extraction module and an aesthetic feature extraction module.We first apply an expansion operation on each text region and output the region of expansion.With the corresponding text anchor,the aesthetic feature extraction module uses the Ro IAlign and Ro EAlign operations to extract the saliency feature and composition feature.In the design process,except for the information explicitly displayed,if some infor-mation can be implicitly embedded into the graphic design,it will be a valuable appli-cation.We present an approach called for embedding information into graphic designs.The designed model considers the salient features of graphic designs to reduce the explicit visual loss caused by encoding.To further support large-scale encoding and decoding,we consider the characteristics of graphic designs and propose a saliency-based QR code lay-out algorithm.We present a variety of practical applications and conduct a comprehensive evaluation of the perceptual quality of encoding,decoding success rate,anti-attack capa-bility,time performance,etc.The evaluation results demonstrate the effectiveness of our method.Overall,the proposed methods,which are based on visual perception technology,can be applied in a variety of actual scenarios with good visual effects.