Fingernail Detection System Using Differences of the Distribution of the Nail-Color Pixels

摘要

As far as we know, previous studies have only used color information for fingernail detection. Images of the palm, however, have a lot of pixels that have the same color values as the fingernail. Other kinds of some information are required to distinguish fingernails from areas of the palm having the same color. We found that there seem to be differences in the distribution of nail-color pixels, which is useful for fingernail detection especially when a camera captures a palm side. We found differences of the areas of the fingernail and skin, and have confirmed these are available for corresponding to the palm. In this paper, therefore, we propose a fingernail detection system using differences in the distribution of the nail-color pixels. The system proposed first fixes provisional nail color indicating nail-likelihood as a calibration process. In main processing, the system has two-step searches. In the first step, it detects all areas which have similar colors to the provisional nail color. Then, in the second step, it distinguishes fingernails from other areas using differences in distribution of nail-color pixels. In the experiments, we investigate the relationship between wrist rotation angles corresponding to the gradient of the nail and percentages of the correct detection. As a result, we confirm that our proposed system can detect only fingernails with at least 85% correct answer or more, except a camera is almost facing to the side of a little finger or thumb.