Griaule Biometrics

首页 » Ridge feature extraction

Ridge feature extraction

The minutia extraction is difficult in low image quality image, therefore, other information such as ridge shape and texture could be used to match the fingerprints, although they are not too distinctive as the minutiae. There has been proposed method that explores the ridge feature. In this subsection is presented the extraction of fingercode.

The fingercode is a vector of fingerprint feature obtained by applying a convolution to the original image with a Gabor filter. In this vector are stored the value of the standard deviation of each region.

The dimension of the Gabor filter is generally 32x32. At least 4 directional filter are enough to extract global ridge information, however, 8 directional filter are necessary to obtain the local characteristics. The figure below shows the result of applying 8 directional filter to the fingerprint image.




Result of convolution of the image with the Gabor filter in eight directions (0°, 25.5°, 45º, 67.5º, 90º, 112.5º, 135º, and 157.5º)