AUTOMATIC MULTILEVEL THRESHOLDING BASED ON TWO-STAGE OTSU'S METHOD WITH CLUSTER DETERMINATION BY VALLEY ESTIMATION

The modified two-stage multithreshold Otsu (TSMO) method based on a two-stage Otsu optimization approach is proposed for multilevel thresholding. The proposed method yields the same set of thresholds as those obtained by using the conventional Otsu method, but it greatly decreases the required computation time, especially for a large number of clusters. In addition, an effective method of histogram-based valley estimations is presented for determining an appropriate number of clusters for an image. Various real-world images were used to evaluate the performance of the proposed method. Experimental results show that the speed of computation for the proposed method is about 19000 times faster than that for the conventional Otsu method when the number of clusters is 7.