Image segmentation method based on genetic algorithm and OTSU

Image segmentation is to extract the significant features or the features to be applied from the image, but different types of images and different applications requirements required different features to be extracted and therefore the feature extraction methods are also different. At present, there is not a universal optimum image segmentation method. Genetic algorithm is a global optimization process of a population. It is more efficient than blind search and it is more universal than the methods for specific problems. In fact, it is a solution pattern irrelevant to the problems. However, genetic algorithm has great uncertainty and premature convergence; therefore, it can be used to solve problems with other algorithms. The maximum between-class variance (Otsu method) is deduced based on the least square method. Its basic idea is to take a certain grayscale as the threshold of the histogram of the image, divide the image into two groups and calculate their variance. When the variance is the maximum, take this grayscale value as the threshold to segment the image. This paper proposes an image segmentation method based on genetic algorithm (GA) and Otsu method. This method finds an optimum solution from the solution space of the image segmentation to maximize the variance and GA can search the optimum solution and the maximum variance quickly. The optimum solution at this time is the optimum threshold of OTSU method. The test experiment proves that the method of this paper can achieve higher segmentation accuracy and reduces its time complexity; therefore, it is a feasible image segmentation method.