인물사진의 감성 선호도 측정 자동화 프로그램 연구

초록

Recently increasing usage of portraits on the web expresses opinion medium of diverse emotions. Our research aimed to develop an automatic program model for measuring preference in terms of subjective, objective and aesthetic aspects. Firstly we analyzed subjective factors of viewers both portraits' contents(positive, neutral, negative) and portraits' color tone(cool, neutral, warm). For the aesthetic image quality assessment factors, we measured and defined face area, composition, ratio, and out-of-focus of background. We measured and statistically analyzed the objective measurement factors; contrast, color, noise and dynamic range. Throughout the main experiment, we proposed Image Quality Assessment Program model and it is adjustted the 10 highly preferred portraits. We ranked numerically 10 portraits as each item and analyzed correlation for the reliability and objectivity between professional's survey and Image Quality Assessment Program model results. With the results, the correlation coefficient was r=.73(p<0.1), the high level of correlation. Those results are the final preference of portrait's viewer in aspects of subjective, aesthetic, and objective assessment factors. As our research represented a novel automatic method image quality program model for measuring emotional preference. we synthesized differently treated research areas; subjective, objective, and aesthetic assessment factors. These results quantitated viewer's preferred portraits assessment factors and it analyzed the preference enhancing factors in order to give the positive effect to related manufacturers and researches. Also, it means that it is possible to measure viewer's emotions to numerical preference. We expect the related researches for estimating viewer's emotion and using it practically.

키워드