Iris recognition by local extremum points of multiscale Taylor expansion

Iris recognition by fusing different representations of multi-scale Taylor expansion

“…The iris segmentation in our work consists of three main steps [24]: (1) localizing the inner and outer boundaries of the iris, (2) the optional detection of occluding upper and lower eyelids, and (3) the detection and removal of reflections on the cornea or the eyeglasses. For iris segmentation we use standard state-of-theart image processing techniques (see also [40] for a general overview on eye detection.…”

“…For the technical details of the segmentation steps we refer the reader to our earlier work [24] where we discuss each step in more detail. Fig.…”

“…The most significant local extrema of the Taylor expansions define the most significant iris texture variations. This method is an extension of the one proposed in [24]. In this work, we provide an efficient matching approach and present extensive evaluation results on recent databases.…”

“…For large-scale applications, it is compact in size, and we provide an efficient and elastic matching scheme, the enhancement of the local extrema-based approach from [24] with an efficient similarity measure, a novel matching scheme that combines the local extremabased approach with the binary feature approach which has (a) a compact representation of the iris data and which is (b) efficiently to compute, evaluation results for all three above approaches on the databases Casia 2.0 (device 1), ICE-1 and MBGC-3l and the recognition results are among the highest in the literature.…”

“…In Sections 3, 3.1 and 3.2 we discuss the binary and the local extrema based iris representations. For better readability, we also briefly summarize the extrema-based approach from [24]. In Section 4 we discuss the elastic matching scheme for the phase-based approach and the efficient matching scheme for the local extrema-based approach.…”