Improvement of face identification by using image sequence (パターン認識・メディア理解)

概要

論文の詳細を見る
We developed 2 novel method to improve the performance of full automatic face identification in video. The face identification system we use is based on global facial featuring such as Eigenface or Fisherface for face identification. Our methods are focusing on improve feature crediblity using face sequence from video. The first proposing method use face sequence tracking to refine automatic facial feature locating result, which will lead to better result for face normalization. The second method use multiple matching between face sequence to improve matching precision. Premilinary experiments result show that the 2 methods can improve the final performance of face identification in video effectively.
社団法人電子情報通信学会の論文
2009-02-12