Reliable iris feature extraction by local thresholding with optimum block size
Ryo Ishikawa, Tomohiko Ohtsuka · 2017
In this paper, we propose the new iris feature extraction method that uses local thresholding with block size fitting to achieve a reliable iris authentication technique. The dispersion index is used to analyze the degree of variation in the pixel intensities. By calculating the pixel intensity variance for the block size, it is possible to quantify the degree of contrast and brightness in the block. The proposed method can adjust the block size for local thresholding by utilizing the pixel intensity variation. Evaluation experiments show that it is possible to extract the iris feature from low-contrast patterns with high accuracy, which was difficult to achieve using the conventional approach. Experimental results show that the EER (Equal Error Rate) of the proposed method is reduced to 0.4%, and the processing time increases to 3.3 ms when compared with the conventional approach.