Examination of Image Retrieval System Using Subjective Image Information

The demand for an efficient retrieval method has increased as the number of digital content increases. However, a versatile image retrieving method does not exist though a lot of researches have been tried to achieve one up to now. This paper reports KANSEI perspective. It orthogonalizes to an existing retrieval method, and contributes to the improvement of retrieval precision by combining it. KANSEI perspective is a method to express image characteristics by extracting subjective features by watching the image. The features used in KANSEI perspective are categorized as follows: the object position in the image, the relative camera position, and the camera elevation. By combining these three parameters, the amount of feature of the image is assumed. The advantage of this method is to achieve less fluctuation of the expression by those who describe it. In addition, the amount of feature expressed by this method is to be described as meta-data item compliant to MPEG-7, and the image retrieval test bed using this method has been completed