Research And Application Of Font Style Migration Algorithm Based On Deep Learning

With the development of digital society,the demand for different font styles is increasing,small daily office,emotional atmosphere in letters;big cultural expression,brand image of an enterprise.And the need for this personalized style font will only increase with the development of society.However,the design of Chinese font is different from the English font.The Chinese font strokes are complex.When designing the new font,professional font designers need to analyze and write the strokes of each word,and then conduct binary coding to make it into a digital font.This process is not only tedious,but it takes a lot of time.Not only that,our country as an ancient country with a long history,with the development of history,the ancients also left many unique style of font,these font,has an extraordinary significance to research the background,however,in the inheritance of history,some font missing happened more or less,some ancient excellent font even only less than one hundred words.Therefore,it is of important research significance to study a font style migration algorithm,used to quickly generate style text,or generate other characters of the same style with a small amount of font style.The main research contents of this article are:(1)Starting with image style migration,it studies the commonly used algorithm model in the field of style migration,and compares the differences and difficulties of font style migration and image style migration.(2)In view of the characteristics of font style migration,the generative style migration algorithm was selected to study the font style migration.On the basis of previous research,a single discriminant font style migration model with WGAN network as the overall structure was realized.Experiments found that the single discriminator model could not retain the font of complex strokes,which leads to the font picture after the migration of style migration,stroke details incomplete,and fuzzy font style.(3)In view of the defects of single discriminator model,a font style migration model based on dual discriminator model was designed,introduced dual discriminator to independently judge the font content and font style,and used WGAN-GP network as the overall structure to solve the defect problem of WGAN network.Finally,the designed dual discriminator font style migration model was used to migrate the font style.The verification results show that the font picture of the dual discriminator model is clear and the stroke details are relatively intact,and the objective analysis of the peak signal to noise ratio(PSNR)and structural similarity(SSIM)through intuitive perception and objective evaluation criteria fully demonstrates the feasibility of the dual discriminator model method designed in this paper.