Iris Recognition Using Wavelet Features and Various Distance Based Classification

This paper proposes wavelet features based iris recognition approach adopting K-nearest neighbor (K-NN) classification with various distance measures. Image pre-processing, feature extraction and iris classification are the three key stages in this proposed approach. Very well-known image enhancement, iris segmentation and iris normalization techniques are employed in image pre-processing stage. The wavelet features are extracted from the normalized iris images by applying various types of mother wavelets like as Haar, Log-Gabor, Morlet, Daubechies, Symlet, Coiflet, Bi-orthogonal, Meyer, Fejer-Korovkin and Reverse bi-orthogonal wavelets. In classification stage, the K-nearest neighbor is adopted as the basic classifier with Chebychev, Jaccard, Minkowski, City Block, Euclidean, Spearman, Correlation, Braycurtis, Cosine and Canberra distance metrics. We conduct extensive experiments on the distance database: CASIA-v4 and the results are demonstrated as the tabular form. The classification performance is evaluated by receiver operating characteristic curves. The empirical results prove that these distance measures influence the system performance greatly and Spearman distance is the reliable metric for K-NN classifier in Symlet wavelet features based iris recognition.