Research of the algorithms preprocessing of biometric images for the purposes of verification

Biometric identification and verification is more often used in problems of access control. In this paper for the purpose of verification as a biometric feature selected geometry of the human face.

For verification system developed module extraction biometric features, which selects the most significant elements of the biometric image and the comparison module biometric samples. As a measure of similarity used similarity coefficient based on a normalized Euclidean distance. To reduce errors comparing the first and second kind (FAR, FRR) investigated the following pre-processing algorithms: rotation, scaling, principal component analysis (PCA).

The simulation showed that the using of different pre-processing combinations allows to reduce the error ratio. The best pre-processing combination: scaling and principal component analysis. Scaling pre-processing is present in all the experiments, the performances of which were better than others. The sequence performed pre-processing virtually no effect on the results.