Fingerprint Preprocessing And Texture-based Feature Extraction Algorithms For Personal Recognition

Fingerprint is the most widely used biometric modality because of its high immutability, individuality and universality. It has been employed for personal recognition in a large number of law enforcement and civilian applications since decades. Fingerprint recognition has achieved high accuracy and is a well matured technology for good quality fingerprints however recognizing low quality fingerprints having low contrast such as dry or wet fingerprints is still a challenging job and need a continues efforts and further research to make the system performance up to the mark, get an error-free and human independent recognition system. In this thesis we have developed an improved framework to optimize the performance of algorithms in automatic fingerprint recognition systems.Accurate segmentation of fingerprint ridges from noisy background is necessary for efficiency and accuracy of subsequent enhancement and feature extraction algorithms. Fingerprint image segmentation plays an important role in the automatic fingerprint recognition system. The traditional methods used for fingerprint segmentation divide the image into non-overlapping blocks, resulting in blocking effects at the foreground edges. We put forward a novel approach to segment the fingerprint images adaptively using the image blocks overlapping method. In addition to this a new post-processing technique is applied to the segmented images to make the segmentation process more accurate and robust. Our experiments show that this method gives better results for varying quality images compared to the methods based on non-overlapping blocks and demonstrate the improved performance of the proposed method.Fingerprint Image Enhancement is an essential process of automated fingerprint identification system. The fingerprint acquired by the scanner is usually of low quality and should be enhanced before the feature extraction and matching processes for an authentic and reliable user identification. Enhancement of fingerprint images improves the ridge-valley structure, increases the number of correct features thereby improving the overall performance of the recognition system. We propose a new method of fingerprint image enhancement using a combination of diffusion-coherence filter and spatial-domain2D-Gabor filter. Additionally the blocks overlapping technique is used to remove the blocking artifact in the enhanced image. The experimental results demonstrate the improved performance of the algorithm in the core region and in the plane ridge-valley pattern of the image compared to the Diffusion and Gabor-based methods if used disjointedly.The texture-based features of the fingerprint have been selected for recognition due to it’s easily extraction over minutiae-based features in low quality images. In this method, the Core point is found initially using Poincare index method. Then the dominant fingerprint region around the core point is selected and enhanced using the diffusion coherence technique for image enhancement. The gray level co-occurrence matrix (GLCM) is applied to this enhanced region to find out the fingerprint most significant statistical descriptors such as energy, contrast, homogeneity, entropy, dissimilarity, maximum probability and variance. Finally the K-Nearest Neighbor (KNN) Classifier is adopted for the recognition of unknown fingerprint images using texture based feature set. The results of our experiments demonstrate the improved performance of the algorithm.We have worked to improve the performance of fingerprint recognition algorithms, by consolidating an "Adaptive Segmentation of Fingerprint Images Using Blocks Overlapping Algorithms","Fingerprint Tmage Enhancement using Coherence Diffusion Filter and Gabor Filter" and a "GLCM-based Fingerprint Recognition Algorithm" together.