Research On Color Emotion Evaluation Of AI Images From The Perspective Of Kansei Engineering

The rapid development of the mobile Internet and the wide application of deep learning technology have provided people with a large amount of visual content,such as videos and images.In order to provide people with better pictures and videos,the requirements for AI and robots to discover,process,and create beauty should also increase accordingly.This paper hopes to propose an AI image color emotion evaluation method based on Kansei Engineering theory,which can help AI to correct the creative direction in the process of creation and obtain pictures that are more in line with human aesthetics.This paper mainly uses the research method of Kansei Engineering to digitize the abstract feelings of images.Firstly,the research used factor analysis and semantic difference method to reduce the dimension of perceptual vocabulary,and obtained three groups of perceptual vocabulary:steady-frivolous,refreshing-dirty,elegantvulgar,which merged into the evaluation dimension of images and characteristic color groups in the follow-up research.Then,three image extraction algorithms are used to extract the main colors of AI images,and Kansei Engineering is used to judge image emotions and color group emotions in three emotional evaluation dimensions.Through the correlation analysis of image emotion and characteristic color group emotion extracted by three algorithms,the conclusion that the color emotion extracted by octree algorithm is closest to the original image emotion is obtained.Through multiple linear regression,the formula of image color emotioncharacteristic color group emotion is obtained.Finally,the extracted RGB colors are transformed into HVC colors,and their order factors and complexity factors are calculated respectively.Through the construction of regression equation,the formula of emotion-order complexity of characteristic color group is deduced,and the formula of emotion-order complexity of image color is finally obtained by combining with the formula of emotion of image color group and characteristic color group,which has passed the experimental test.