This type of method is based on the integration of basic image statistics such as mean, variance, entropy, histogram, and multidimensional space distance to form new high-order statistical features for image segmentation. Commonly used image high-order statistics include image gray level co-occurrence matrix, entropy and fuzzy entropy of two-dimensional gray level histogram, maximum correlation criterion of two-dimensional histogram, etc. These algorithms often have high complexity and a huge amount of computation, and it is necessary to fully combine genetic algorithms, immune algorithms, neural networks and other optimized numerical algorithms to obtain image segmentation results.