著者

矢野 雅之 勝間 大輔 清水 大輔 渡邊 睦

出版者

一般社団法人電子情報通信学会

雑誌

電子情報通信学会技術研究報告. PRMU, パターン認識・メディア理解 (ISSN:09135685)

巻号頁・発行日

vol.105, no.674, pp.149-154, 2006-03-10

---

画像認識を行う際、まず領域分割を行って認識対象の侯補となる領域を設定することが行われる。しかし、領域数は一般に未知であり、又、明度/色相値の類似性のみで領域を生成した場合には影や模様の影響により適切に設定できないという問題がある。そこで本稿では、明度による自動領域分割に対してDFT処理した領域をクラスタリングした分割結果を統合することにより、領域分割結果の精度向上を行う方式について提案する。屋内シーン、屋外シーンにおいて本手法の有効性を確認した。

著者

安陪 光正 矢野 雅之 熊本 敏子 手柴 博子 岡田 くにえ 高岸 達也 山本 厳雄 武市 昌士 安岡 文恵

出版者

一般社団法人 国立医療学会

雑誌

医療 (ISSN:00211699)

巻号頁・発行日

vol.18, no.3, pp.148-153, 1964-03-20 (Released:2011-10-19)

参考文献数

7

---

In the summer of 1960 a girl with cerebral paragonimiasis was admitted to the psychiatric service of Chikushi National Hospital from Kami-agata Country (Tsushima Island), Nagasaki Prefecture and she died shortly after hospitalization. Her parents also were suffering from paragonimiasis, and had no knowledge about the disease, Since this incident, we have carried out epidemiologic studies for four years, and some campaign of education regarding paragonimiasis in Tsushima, Namely, parasitologic studies on the crabs, or the skin test, and explanations concerning life cycle of paragonimus using many slides for the inhabitants at various places have been done. In addition to these, since 1961 in every year 2, 000 copies of the pamphlet for education have been distributed to the people in Tsushima. On the other hand through the newspapers, radio and television the importance of prophylactic measure has been emphasized to the inhabitants, According to our studies, parasitic rates of metacelcaria in the Nita River (especially 24.3% in Kurusu), 42.5% in the Mine River, 59.8% in the Kusu River, 58.1% in the Are River, and 80.0% in the Azamo River, which appeared to be high positive rates.High positive results of the skin test obtained from school boys and girls using the VBS antigen were 22.7% in Sago area, 37.6% in Hamagusu, 21.9% in Are, and 23.8% in Naiin. High positive skin test corresponds with high parasitic rate of metacelcaria in the crabs in these districts, In the skin test for 1816 individuals among the inhabitants in Sago, Nita, Mine and Are areas, where appeared to show high infectious rate, the positive rate was 22.6%. In 1962, the questionnaire was sent to every home in order to know the inhabitants' response to our campaign of education and some following interesting data were obtained.The spread rate of knowledge about the disease among the inhabitants are showing remarkable increase, namely 33%, 34% in 1960, 1961, and 1962, respectively.As to the processes obtained the knowledge, 35 per cent of the people owed to the activities of our group, 34% from the talk of folks with this matter, and 2096 from information by the newspapers, radio and television.In Tsushima, the most families (93 per cent) have tasted the crabs and in 1960 48.2% of the inhabitants ate the crabs. Year by year, however, the eating ratio of crabs was decreasing, i. e., it was 21.9% in 1961, and in 1962 it dropped to only 2.9%. As to the cooking ways of crabs, 66% of people took Miso soup containing grinded crabs, but no one showed such habit as eating the raw crabs.Since 1960 111 carriers of paragonimus eggs have been found, among them 74 instances by us, 35 cases by the Institute of Endemic Diseases, Nagasaki University, and 9 by doctors in Tsushima. In 1963, under cooperation with the National Tsushima Hospital, group treatment for paragonimiasis cases with Bitin was made and 50 volunteer cases have completed the treatment.

著者

矢野 雅之 勝間 大輔 清水 大輔 渡邊 睦

出版者

一般社団法人情報処理学会

雑誌

情報処理学会研究報告コンピュータビジョンとイメージメディア(CVIM) (ISSN:09196072)

巻号頁・発行日

vol.2006, no.25, pp.321-326, 2006-03-17

---

画像認識を行う際、まず領域分割を行って認識対象の候補となる領域を設定することが行われる。しかし、領域数は一般に未知であり、又、明度/色相値の類似性のみで領域を生成した場合には影や模様の影響により適切に設定できないという問題がある。そこで本稿では、明度による自動領域分割に対してDFT処理した領域をクラスタリングした分割結果を統合することにより、領域分割結果の精度向上を行う方式について提案する。屋内シーン、屋外シーンにおいて本手法の有効性を確認した。The first stage of object recognition is generally the region(description) creation to identify candidates of recognized objects by means of segmentation. However, several difficulties exist in the segmentation, such as, unknown region number and miss-segmentation due to shadows or irregular texture patterns. We propose a unified method of both intensity-based segmentation result and texture-based one to improve the accuracy. The texture-based segmentation is executed by clustering DFT outputs. Experimental result have shown the effectiveness of the proposed method.