Fingerprint image quality measurement algorithm

We present a human-visual-perception-model-based algorithm for determining fingerprint image quality that results in a single score to represent the image's quality level. This method would allow an operator who is responsible for the collection of fingerprint images in the field to render, accept, and reject decisions quickly based on calculated quality scores. This method first identifies the fingerprint image's region of interest (ROI) and then targets that area for quality measurement. We then propose the ROI be reduced by 2 in both the horizontal and vertical axes by using a 5 × 5 low-pass filter with Gaussian weighting coefficients. The quality is then determined by the majority orientation within each cell of the image, which is formed by a 9 × 9 pixel block. An image's overall single quality score is calculated by taking the average of all the cells' quality levels.