Classification of facial features for recognition

摘要: A facial feature classification technique that independently captures both the geometric configuration and image detail of a particular is described. The first extracted by fitting deformable template to shape (for example, an eye) in image. This information then used geometrically normalize such way attains standard shape. normalized classified terms set principal components previously obtained from representative training images similar features. stage yields representation vector which can be for recognition matching alone without complication changes expression. Implementation system described results are given its application test faces. These show features reliably recognized using vectors obtained. >