"Kansei" image retrieval system for street landscape-discrimination and graphical parameters based on correlation of two images

Abstract

We define "Kansei" as discriminated subjective interpretation. The paper concerns a theoretical framework for a "Kansei" model and an application of a retrieval system for street-landscape images by using the "Kansei" model. The model is based on the relationship between adjectives and graphical parameters, so users' subjective interpretation is reflected in the retrieval system. We introduce an auto-correlation model which can not only represent direction parameters in images but also is not dependent on a color average. The auto-correlation model can be defined as a correlation between reference area and spatial shifted areas. The operation is similar to that of the optical flow method. We conclude that the independence of graphical parameters is necessary for the "Kansei" model and "Kansei" is based on a process of discriminating our environment by using physical parameters, independent of each other.