Performance Analysis Comparative Study of Fingerprint Recognition Systems

Abstract

Forensic applications, such as criminal investigations, terrorist identification, and national security, require a strong fingerprint identification system. This paper proposes four methods, namely, canny filter, Gabor filter, dual-tree complex wavelet transform (DTCWT), and principal component analysis (PCA), to obtain a high fingerprint recognition rate. Frequency domain filtering is used to enhance fingerprint images. In canny filter, feature extraction based on the gray level co-occurrence matrix (GLCM) is computed from the gradient and coherence images. Fingerprint features are extracted and stored through the eight different orientations of Gabor filter. The redundancy and shift invariance of DTCWT is useful for obtaining high-resolution images with preserved edges. PCA is used to extract the statistical features of fingerprints by reducing their dimensions and complexity. The proposed methods improved the efficiency of fingerprint recognition by combining GLCM-based feature extraction with a K-nearest neighbors classifier. Co-occurrence matrices are used to extract features from the fingerprint image because they are composed of regular texture patterns. The proposed methods increased the recognition rate and reduced complexity and time. The false accepted rate, false rejected rate, and total success rate were improved by the proposed methods compared with those of existing algorithms.