Ordinal Measure of Discrete Cosine Transform Blocks for Iris Identification

Currently, a common method for identifying a person is by means of an identitycard (ID) or combination of an ID and password. The approaches are not very reliable, since the ID can be stolen and password can be forgotten. A more reliable identification system is required. In the last decades, identification systems based on biometrics have been gaining attention, since they are more reliable. Biometrics-based devices identify people based on their physical or psychological characteristics, such as palmprints, fingerprints, gait and iris. Unlike fingerprints or palmprints, irides features distribute randomly, and the features were unique; the features between right and left eyes aredifferent, as well as between twins. Therefore, in addition to reliability, the use of irides can enhance identification accuracy. Purpose of the paper was to improve identification rate of an iris identification method, using ordinal measure of Discrete Cosine Transform (DCT) coefficient. The input iris image was tiled into blocks of 8x8 pixels, then the DCT was applied to each blocks. The AC coefficients of each block were sorted from the smallest to the largest values, in which the sorted values were referred to as ordinal measures.Identification was accomplished by measuring a distance between the ordinal measure of the input images with the ones of the existing images in the database using Minkwoski distance metric. Proposed method increased the averaged identification rate as compared to the previous method by nearly twice from 33% to 61.4%.