Kansei factor space classified by information for Kansei image modeling

Abstract

Now that intelligent computing era is coming, it becomes necessary to retrieve and process images according to human Kansei or preference. Application and research on Kansei information of image, especially, are mainly focused on satisfying individual taste. However, existing researches are focused on content-based retrieval such as low-level information and visual features that make it difficult to retrieve and process images according to human high-level Kansei information. Therefore, we suggest solving this problem by defining relations between visual information and Kansei and establishing “Kansei Factor Space” using Kansei vocabulary. The Kansei factor space defines the systematical Kansei information, a Kansei modeling, by use of its vocabulary and represents Kansei of each image. In addition, through Kansei factor space, Kansei can be measured in terms of quality. Thus, our study could serve as a basis for Kansei information image processing.