Fingernail Detection Method from Hand Images including Palm

We think that hand pose estimation technologies with a camera should be developed for character conversion systems from sign languages with a not so high performance terminal. Fingernail positions can be used for getting finger information which can’t be obtained from outline information. Therefore, we decided to construct a practical fingernail detection system. The previous fingernail detection method, using distribution density of strong nail-color pixels, was not good at removing some skin areas having gloss like finger side area. Therefore, we should use additional information to remove them. We thought that previous method didn’t use boundary information and this information would be available. Color continuity information is available for getting it. In this paper, therefore, we propose a new fingernail detection method using not only distribution density but also color continuity to improve accuracy. We investigated the relationship between wrist rotation angles and percentages of correct detection. The number of users was three. As a result, we confirmed that our proposed method raised accuracy compared with previous method and could detect only fingernails with at least 85% probability from -90 to 40 degrees and from 40 to 90 degrees. Therefore, we concluded that our proposed method was effective.