A multi-threshold image segmentation approach using state transition algorithm

Thresholding is an important approach for image segmentation and analysis. In this study, the combination of normal distribution functions is used to fit the normalized histogram of the original image since each normal distribution function represents a pixel class. On the other hand, state transition algorithm (STA) is a promising method for solving complex optimization problems. By transforming the fitting problem into an optimization problem, the STA is used to select the optimal parameters in the fitting function. Experimental results of several images show that the proposed approach is efficient and effective for multilevel thresholding problems. Comparisons with OTSU, PSO and GA also demonstrate that STA not only outperforms computationally efficient but also provides competitive thresholding results.