Colorful Natural Scenes Retrieval based on Affective Features Hierarchical Model

From 1990s, with the genesis and rapid developments of Kansel Engineering and Affective computing, affective information processing methods have been used to analyze and process images to extract affective features contained by images. Image retrieval by affective features attracts more and more attentions in recent years. Colorful natural scenes are a kind of images with remarkable characteristic in evoking viewers' subjective experiences, which always leave deep impressions on viewers, evoke strong emotions on them, and even put them under some longstanding feelings. Indexing scenes by affective clues may describe images more completed and provide users a more friendly way to touch them. This paper mainly discusses issues on features modeling, trying to establish a common model for affective features from the aspect of information retrieval. Firstly, it gives a brief review on related research. Next it analyzes and points out the hierarchical characteristic of affective features, from which a model named AFHM is derived. Then it interprets AFHM in detail. After that, it carries out two experiments with colorful natural scenes to explore some disciplines when affective features are used as indexing descriptors and searching queries. The findings reveal the further characteristics of affective features in AFHM. Finally, it summarizes the conclusions.