Noise removal is a very important step in an iris segmentation process. Iris regions are usually occluded by Eyelid and eyelashes. For overcome this problem, we present a robust method for eyelid and eyelashes segmentation based on wavelet transform. Our approach follows two main stages. First, eyelashes are removed using wavelet transform. Then eyelids boundary are modeled with a parabolic curve. Second, Eyelashes are modeled by Hough transform. Afterwards eyelashes are segmented using neural network. Experimental results on a set of 756 images show that the accuracy of proposed method leading to accurate eyelid and eyelash segmentation.