A new segmentation method for non-ideal iris images using morphological reconstruction FCM based on improved DSA

Abstract

In any accurate iris recognition system segmentation of iris plays a vital role. The noise, specular reflections, eyelid/eyelash obstruction, and intensity inhomogeneities in an image make the segmentation more difficult. In this paper, a novel technique is proposed to segment the iris from images that are taken under uncooperative image conditions. The proposed method segments the image in two stages. Firstly, Morphological reconstruction fuzzy c-means clustering (MRFCM) based on an improved differential search algorithm is implemented before the segmentation step. The MRFCM can preserve image contours even in the presence of noise. Secondly, the iris is isolated from the undesired regions of an eye image by implementing geodesic active contours driven by a modified stopping criterion on the resultant images of the pre-segmentation step. The accuracy of the method presented has been tested on the databases such as CASIAv3-Interval, UBIRISv1, MMU1, IITDv1, and MICHE-I. The segmentation accuracy has been demonstrated and compared with other existing methods present in the literature. The obtained results are promising and the proposed model is outperformed the existing methods.