关键词: 风景图片, 情感特征, 颜色直方图, 多元线性回归

Abstract: The characteristics of image affective features were analyzed and a method was provided to divide affective features into 3 levels. Typical affective features for colorful natural scenes were chosen and an investigation was carried out to collect users' impressions on images. Then, based on color features and users' evaluations, the mapping between color and subjective impressions was established by multiple linear regressions, which could be used to extract affective features automatically. Finally, the validity of 3 levels was verified. Besides, the mapping is also testified effective to forecast and index the affective features correctly.

Key words: natural scenes, affective features, color histogram, multiple linear regression