Bulletin of Computational Science Research Center, Hosei University

固有パターン

有限差分

Dynamic Image Cognition

Eigen Pattern

Finite Difference

Most of the conventional image cognition methodologies are based on the geometrical singular points extraction of the target images． In human face image cognition， it is proposed that the method of discriminating the ratio of the distance between eyeballs and a mouth etc． as characteristics of each face information． This recognizing method has problems such as a definition or extracting method of the singular points， and change of the singular point distribution by change of facial expression． In order to overcome these difficulties， we have previously proposed a new image cognition methodology based on the color information extracting strategy． In the previous paper， we have reported that the image cognition method based on the color information which composing each of the target images could recognize in fairly good accuracy． Furthermore， it has been reported already that this method is generalized to the dynamic images cognition． As a result， it has been reported that the dynamic images was possible to recognize more precisely than static image． In this paper、we propose the method of extracting the time variations of the color information by means of the finite differences． It is shown that this new methodology makes it possible to cognize the target objects having any background images．

208 情報メディア教育研究センター ＞ 紀要 ＞ 法政大学計算科学研究センター研究報告 ＞ 18