Determination of the reference point of a fingerprint based on multiple levels of representation

In this article we propose a technique to detect the reference point of a fingerprint based on the main supposition that the crests belonging to the superior area of reference point usually have a orientation quantized of 0 degrees. To take advantage of this fact it is accomplished an analysis of the orientation image considering their multiple levels of representation in relation to their values of quantization. This is a new approach that allows the use the morphologic characteristics of the quantized orientation image for each value of quantization. The technique presents some important characteristics: adaptability according to the quality of the image and the possibility of detecting the reference point of the fingerprints A and TA with acceptable error rate for these type of fingerprint. The technique is tested in the whole database of DB4 of NIST getting an estimation error of the reference point of 8.32% for a total of 4,000 images.