Algorithms for reconstructing fingerprints

In the field of forensic science, fingerprints are the most important evidence for personal identity. The principle of exchange established by Locard governs it. Fingerprint identification is the most effective approach for minimising concerns about the identity of fingerprints when comparing them with the Automated Fingerprint Identification System (AFIS). Investigators acquire three sorts of prints from the crime scene: patent fingerprints, latent fingerprints (also known as incomplete fingerprints), and plastic fingerprints. The analysis of the incomplete fingerprints is a challenge for fingerprint experts. Various algorithms techniques have been developed to make the work of fingerprint experts easier and more effective. These algorithms are mostly based on the second and third level characteristics.

Shally Chauhan and Vinny Sharma focus on the numerous algorithm strategies used for fingerprint image reconstruction, how these methods are employed, what their process and accuracy are, and the faults in this review study. When the minutiae were removed from the stand-up case, the ridge frequency field, and the minutiae distribution, the reconstructed fingerprints were very close to the original fingerprints. However, it still works in some cases, including false images for analysis.

Fingerprint impressions have characteristics that are typically a combination of ridges and grooves in the impression. Fingerprint verification has traditionally relied on characteristics that are divided into three distinct levels. From these three layers of depth within the print, a variety of details are revealed.

First level: The overall flow of the friction ridges, the types of patterns, and the position of the core and delta are all first-level details. This provides a general flow direction for each print that will be analysed. In first-level details, creases, scars, and other deformities are investigated by examining the directions and locations of the features.

Second level: It includes the minutiae’s types and relatives’ positions within the overall pattern. The beginning of the ridge, the ridge path, and the point where the ridge flow ends are all included. The true path is traced in the case of second level details. All of these details are more precise to where a ridge comes to a halt, bifurcates, converges, or diverges.

Third level: The intrinsic or congenital ridge forms, which include edge shapes, sizes, pore shape, and relative pore placements, are discussed at the third level. In terms of sequencing, the third level details are unique. The condition for this level is that it does not exist without the first and second levels. It aids in the examination of pore position and ridge shape.

It is now possible to reconstruct a fingerprint impression image from a sample, which can subsequently be compared to the original fingerprint image with excellent accuracy. Grayscale, segment, skeleton structure, and small image are the four types of schematic representation approaches used by fingerprint comparison systems. For this objective, a variety of algorithms techniques are accessible.

  1. Convolutional Auto Encoders Neural Network
  2. Amplitude Modulated and Frequency Modulated
  3. Type 1 Attack and Type 2 Attack
  4. DORIC (Differentiation of the Orientation Values Along a circle)
  5. Orientation Field Modelling
  6. Sparse Auto Encoder Algorithm

The image of the first fingerprints of the given input, i.e., set in input, is obtained by reconstructing their appearance. The following are the three most important reasons for doing so:

(1) To show the requirement for a minutiae template,

(2) To increase the interoperability of fingerprint templates created by diverse sensor and algorithm combinations, and

(3) To construct fingerprint combinations.


Shally C, Vinny S. Fingerprint Reconstruction using Algorithms. J Forensic Sci & Criminal Inves. 2022; 15(5): 555922. DOI: 10.19080/JFSCI.2022.15.555922

error: Content is protected !!