6-9 バイナリ局所特徴を用いた画像検索の精度改善(第6部門 メディア工学(画像認識))

抄録

In this paper, we propose a new image retrieval method based on binary local features and Bag-of-Visual Words framework. We extract uncorrelated portion from bit strings of feature vectors in the same cluster in order to efficiently represent the distance between the vectors. The effectiveness of the proposed method is confirmed in experiments.