A novel iris recognition technique using monogenic wavelet phase encoding

Encoding the local phase of the iris texture has come out as a promising way of generating extremely efficient feature vectors. Gabor wavelets yield a local phase representation distributed over several scales and orientations. One of the very promising attempts at obtaining a useful phase information is from the field of analytic signals and wavelets. The monogenic signal is an extension of the analytic signal to multiple dimensions. In this paper, we propose a novel technique that utilizes multi-resolution monogenic phase for iris recognition. The motivation behind the proposal has been discussed and supported by experiments. Here we suggest that the distinctive ability of the monogenic wavelets to simultaneously extract local phase and orientation can be exploited for iris recognition applications. In order to test the effectiveness of the proposed method, the reconstructions obtained using multi-resolution monogenic phase are compared with the Gabor and Fourier counterparts. Improvement observed in the structural similarity index (SSIM) value reveals the suitability of the monogenic wavelet phase for the applications involving texture identification. Although monogenic wavelets can be used for general texture identification, here, we specifically consider the iris recognition application. Experimental results obtained on the CASIA version 1 and the version 4 interval databases show that the monogenic wavelets can achieve an accuracy comparable to that of the Gabor wavelets.