Multi-resolution Face Recognition System Incorporating Fuzzy Integration

概要

論文の詳細を見る
Face recognition has proven to be a difficult task for computers. Methods based on location and mapping of facial features are generally considered to have insufficient information for reliable matching of large data sets and neural network methods require large training sets and CPU intensive training phases. General image processing techniques have been used to recognise faces. However, these methods are very sensitive to the registration and scaling of the input images. We utilise a method which does not suffer from these disadvantages. By employing the fuzzy integration of results at different image resolutions, we are able to obtain recognition rates of 97.9% at false access rates(unknown persons reported as recognised)of 2.43% on a database of 11600 images of 116 persons, 60 of whom are assumed known.
社団法人電子情報通信学会の論文
1995-03-27