LAUSR

Abstract Binarization of fringe patterns is a commonly used process in the fringe skeleton method for optical fringe pattern analysis. Most of the existing binarization algorithms of fringe patterns are based on the determination of gray level threshold. However, spatial intensity inhomogeneity induced by the uneven illumination makes the gray level threshold hard to determine. In this paper, we present a method for the binarization of fringe patterns based on a modified Fuzzy c-means clustering (FCM) algorithm. This method permits the clustering of pixels influenced by their immediate surroundings. The pixels in the fringe pattern are clustered into white fringes and black fringes without calculating thresholds. We evaluate the performance of this method by applying it to computer-simulated and real fringe patterns. Experimental results demonstrate that the proposed method can achieve the desired binarization results and perform better than the previous binarization methods, under non-uniform illumination conditions.