図形中の目立つ部分を考慮した類似図形検索

概要

論文の詳細を見る
This paper has presented a feature considering prominences in binary images. The regions that could attract humans have been examined by questionnaires and the measurement for path of human eyes using the eye mark recorder. From these results, the prominence in binary images is represented as a part which is enclosed by a region of opposite color, a part whose shape or size is extremely different from others. From the measurement of the path of human eyes, since the base point of the path is very close to the centroid point in the image, the origin of the coordinate axes is regarded as the centroid point. Adopting the results obtained from humans shown above, the prominences are represented as third-dimensional vectors. The proposed feature has represented a degree of prominence in binary images. In this paper, using the proposed feature, experiments of image retrieval have been conducted in order to show the efficiency of the feature in similarity retrieval of images. Therefore, the proposed feature is effective to a preprocessing of the classification and retrieval for binary images.
工学院大学の論文