A comparison of fingerprint enhancement algorithms for poor quality fingerprint images

Image enhancement is a common step in most fingerprint feature extraction algorithms. Unfortunately, many fingerprints are poor in quality, which makes extracting reliable minutiae difficult. There are many factors why this may be the case. There might be too much noise or ghosting on the image, or damages (such as scars or creases) on the fingers itself caused by the person's line of work, such as secretaries who deal with a lot of paper in their job, or people who perform manual labor as their occupation. This research attempts to evaluate the effectiveness of image enhancement by comparing three different algorithms, including the use of power transformation in the frequency domain, smoothing on the spatial domain and contextual filtering using Gabor Filters. The experimental results definitely showed improvements after enhancing poor quality fingerprint images, especially when the image is processed in the frequency domain. Contextual filtering also works well in enhancing images based on data in the local context, but in order for it to be more effective and be able to construct better enhanced fingerprint image results, it should also be accompanied with data in the global context.