Feature Extraction of Iris Based on Texture Analysis

In general, a typical iris preprocessing system includes image acquisition, quality assessment, normalization and the noise eliminating. This paper focuses on the middle issue and describes a new scheme for iris preprocessing from an image sequence. We must assess the quality of the image sequence and select the clear one from this sequence to the next step. After detecting the pupil coarsely, we get the radius and center coordinate. We can extract local texture features of the iris as our eigenvector, then utilize k-means clustering algorithm to classify the defocused, blurred and occluded image from clear iris image. This method obviously decreases the quality assessment time, especially some people's iris texture are not distinct. Experiments show the proposed method has an encouraging performance.