Frequency filtering to suppress background noise in fingerprint evidence: Quantifying the fidelity of digitally enhanced fingerprint images

Abstract Fingerprint evidence can benefit in image quality if transformed using digital image processing techniques. This is especially true when considering prints that cannot be easily lifted (such as those deposited on porous paper substrates), or when the mechanism of lifting does not effectively reduce background interferences. In these instances, frequency filtering is one type of mathematical transformation that can serve to increase image clarity and the ability to extract minutiae relevant to pairwise comparisons. To quantify the impact of frequency filtering on image quality, high quality and low quality (noisy) prints were collected. The high quality prints served as exemplars that were compared to the low quality prints both pre- and post-filtering. The resulting pairwise match scores indicate that: (1) frequency filtering has a low probability of creating false positive associations, (2) 90% of the post-filtered images result in a normalized gain in match score, (3) frequency filtering doubled the probability of obtaining match scores greater than 30% (for the automated algorithm employed in this study), and (4) filtering can double the probability of obtaining 10 or more matching minutiae when comparing same source prints. Overall, the research indicates a reasonable and quantifiable payoff in increased clarity, matching minutiae and pairwise similarity for post-filtered images when compared to known-match exemplars.