Recoloring Artistic Image For Diversity Representation

Artistic images often exhibit unique imagination and creativity of designers,and the color expression also reflects the designer’s emotions.Manually modifying the color scheme of artistic images requires a huge amount of work,therefore researchers have begun to explore automatic colorization methods based on color harmony relationships.In existing image colorization methods,the extraction of the color palette mostly relies on user interaction or fixed values of k-means,which makes it difficult to accurately reflect the color of the image.Therefore,this paper proposes a method based on adaptive color palette,which can better reflect the color richness of the image.To better serve designers,this paper also constructs an artistic image color relationship model to automatically extract color information from artistic images,interpret and analyze their inherent correlations,such as color form,luminosity contrast,hue contrast,and color area ratio,and organize them into a dataset.Based on this,the paper proposes a region-based and color harmony-based colorization method that outputs recommended color schemes to automatically colorize the image and present the most appropriate color results.During the colorization process,color modifications need to be made while ensuring the content of the image and ensuring that the modified image regions are semantically valid and visually natural.The research work of this paper includes three aspects:(1)Construct an image colorization method based on an adaptive palette to reduce user interaction.This method can obtain an adaptive palette of the image and represent the colors in the image as a linear combination of some basic colors,decompose the image into layers,and output a palette that more accurately reflects the color richness of the image.(2)Construct an artistic image color relationship model to provide color analysis services.The model can automatically extract color information from artistic images,interpret and analyze their inherent correlations,and construct a color relationship model to provide specific color features and color logic for designers.(3)Construct a region-based and color harmony-based colorization method to output recommended color schemes and achieve automatic colorization of the image.This method captures the color features in the dataset by training on the image dataset and stores color features that have color harmony relationships.Based on the color feature relationship to be colorized,the recommended color scheme is obtained,thereby achieving automatic colorization of artistic images without user interaction and presenting the most appropriate color results.