Modified OTSU Algorithm Using Quartered Search Method

Abstract

In order to reduce the computational complexity of the OTSU algorithm, in this paper, improved OTSU algorithm using quartered search method has been proposed. When compute the optimal segmentation threshold, divide the whole gray level into four regions, and then determine the region of the optimal segmentation threshold according to the Monotonicity of maximum classbetween variance. The proposed algorithm avoids calculating the maximum class-between variance traversing the whole grayscale, so we can find the optimal segmentation threshold only traversing the quarter region of gray level; it overcomes the shortcoming of huge calculation by using Otsu algorithm to search optimal threshold value. Experimental results show that this method can find reliable optimal threshold value, and improve the processing speed. It is easy for improved Otsu algorithm to meet real-time processing requirements.