Masaaki Fujiyoshi

Improved Iris matching technique using reduced sized of ordinal measure of DCT coefficients

This paper presents an ongoing research on the use of ordinal measure of discrete Cosine Transform (DCT) coefficients as a feature for iris recognition. We proposed to reduce number of DCT coefficients from each 8×8 DCT block that is used to form ordinal measure. The aims were to increase matching rate while minimizing feature size. Four simulation rounds were conducted using CASIA database, each with different coefficients, namely 48, 32, 16 and 8 AC coefficients. It turned out that using as many as 8 AC coefficients from each DCT block resulted in a higher matching rate than using other number of AC coefficients of the block. The proposed method can increase averaged matching rate as much as 5% when using 8 AC coefficients. Furthermore, the proposed technique can reduce the feature size by approximately 80%.