Latent Fingerprint preprocessing: Orientation field correction using region wise dictionary

Abstract

Latent Fingerprint Images have been extensively used by law enforcement agencies in investigating the crime spot and use the necessary information obtained as evidence to validate the criminal in Court. Although an important breakthrough in this direction has already been made in plain biometrics recognition, still identifying biometric such as Face in CCTV footage and Latent Fingerprint in uncontrolled, uncooperative, and hostile environment is an open research problem. Poor quality, lack of clarity, absence of proper mechanism make the latent fingerprint preprocessing problem one of the persistent and challenging problem to extract the reliable features. Dictionary based learning technique has given significant result, in contrast to conventional orientation field estimation methods by reconstructing orientation field to enhance the latent fingerprint image. Distorted orientation field is corrected using orientation patches of good quality fingerprint from region wise dictionary. This paper proposes a fresh idea to construct the dictionary by region wise to correct the orientation field in latent image. To verify the accuracy of enhanced image, a statistical observation has been done and got the promising results. This study concentrates on latent fingerprint preprocessing module towards reliable and efficient (optimal) Latent Fingerprint Identification.

Survey on the impact of fingerprint image enhancement

“…Wang and Liu [64] and Liu et al [65] proposed to use dictionaries of ridge like patches. Kumar and Velusamy [66] proposed to learn dictionaries of orientation patches for later use during enhancement. Schuch et al [67] proposed to train and apply deconvolutional auto-encoders for fingerprint enhancement.…”

“…[65] proposed to use dictionaries of ridge like patches. Kumar and Velusamy [66] proposed to learn dictionaries of orientation patches for later use during enhancement. Schuch et al .…”

A Latent Fingerprint in the Wild Database

“…Yoon et al first developed a latent fingerprint enhancement method based on coarse orientation field estimation and Gabor filtering [28], and an optimized version was then proposed by using Short-Time Fourier Transform (STFT) and Randomized-RAndom SAmple Consensus (R-RANSAC) algorithm to estimate the orientation field [29]. Kumar and Velusamy [30] designed a latent fingerprint enhancement algorithm can correct the orientation field using a look-up table. The latent fingerprint orientation problem has been seen as a classification task by Can and Jain [31].…”