Fingerprint image classification by core analysis

Fingerprint classification algorithms that use both core and delta information are not suitable for the images captured from the general fingerprint input device because the image size is usually so small that the delta points are frequently excluded. The paper describes a fingerprint classification algorithm that uses only the information related to core points. The algorithm detects core point candidates roughly from a directional image and analyzes the near area of each core candidate. In this core analysis, false core points made by noise are eliminated and the type and the orientation of core point are extracted for the classification step. Using this information, classification is performed. The algorithm was tested on 730 images and classification accuracy of 91.6% for the four classes (arch, left-loop, right-loop, whorl) is achieved.