A Fast Decomposed Three-dimensional OTSU Algorithm Based on Cuckoo Search for Image Segmentation

Image threshold method is an important technique for image segmentation. The maximum between-class variance of pixel (Otsu) algorithm has been widely applied in the literature. However, the original Otsu method for image segmentation is very time-consuming, and the segmentation results are often unstable for the image with low signal to noise ratio (SNR). In this paper, a fast image segmentation method (D3OTSU-CS), decomposed three-dimensional Otsu based on the technology of the cuckoo search (CS), is proposed. The proposed method starts by overcoming the complex computational by decomposing the original three-dimensional Otsu into a one-dimensional Otsu and a two-dimensional Otsu. The cuckoo search algorithm is employed to find the optimal threshold vector by the global Levy flight searching, and the between-class variance of the two-dimensional Otsu is investigated as fitness functions. The Experimental results are illustrated to show that the computation time efficiency of the proposed method is increased by about 98.6% than the 3OTSU. In addition, the stability and reliability of the segmentation results by the proposed method outperform 2OTSU and 3OTSU.