構造化モデルを用いた人型商標図形の分類

概要

論文の詳細を見る
In this paper, we describe a method of extracting human-shaped trademark images from trademark images by using structured models. The structured model is obtained by superposing a skeleton of the subject image and a large region existing within the silhouetted subject image. In order to determine the degree of human-like quality of an image, which a person feels about the image when viewing the image, we prepared a questionnaire showing plural images, which were variations on an image of a human-shaped basic image. Since the analyzed results of responses to the questionnaire show the human-like quality of each portion of each image, we define this as the human shape index. It is shown that, by judging the human-like quality of structured models by using the human shape index, human-shaped trademark images can be extracted from trademark images.
日本感性工学会の論文