JIP_V2_N4_RP4 Atluri Srikrishna B. Eswara Reddy M. Pompapathi Journal on Image Processing 2349-6827 2 4 30 38 Regularization, Canny Edge Image, Test Statistic, Figure of Merit, Kernel Identification of edge pixels of the noisy signal without doing regularization is still a challenging problem for researchers. There exists different methods and each method has its own assumptions, advantages, and limitations. Author propose a Non Linear Robust Edge Detector (NLRED), which uses n x n window in order to detect the edges of all possible orientations at noisy images. The proposed method partitions the neighbors of the pixel which is under observation for edge candidature into two sub regions based on differences in the local gray level value. The proposed method calculates test statistic for pixel of each sub region by calculating mean, the placement of each member, index of variability and test statistics. The test statistics with maximum value is considered based on two sub regions. These statistics are calculated for eight different orientations. Among these, the statistic with minimum value is considered for edge candidature. The performance is measured in terms of Figure of Merit (FOM) to show the efficiency of the proposed method by comparing with statistical edge detector CANNY approach, in order to detect all possible edges. October - December 2015 Copyright © 2015 i-manager publications. All rights reserved. i-manager Publications http://www.imanagerpublications.com/Article.aspx?ArticleId=3688