ユーザの意向に適応する多入力画像検索

抄録

In this paper, a new approach for image retrieval that is adaptive to user's intention is proposed. Because image content has so many interpretations to represent by a word, CBIR has been developed as a powerful tool for image retrieval. The conventional CBIR searches and sorts images where retrieved image is similar to a query input image. The retrieval works well when features of appropriate query input image are extracted. However, it is difficult to prepare the optimal feature as well as query image that corresponds to user's desired still image and/or video content.

To solve the problems, a new similarity measure using score function for multi-input query is proposed. Further, an iterative algorithm to adapt user's intention with feedback structure is implemented as user friendly interface in retrieval process. To show the effectiveness of our method, various image retrievals using mixed still image and video were executed. The precision rate was calculated to evaluate the performance of image retrieval. High performance image retrieval was achieved compared with the conventional image retrieval approach.