The image segmentation results under the maximum interclass variance criterion have the problems that the original information is not enough, the real-time performance is poor, the number of iterations in the Pulse Coupled Neural Network (PCNN) model is difficult to determine. In order to solve the problems, a new automatic image segmentation algorithm was proposed based on the fusion of group intelligent algorithm optimized OTSU-entropy (OTSU-H) and PCNN. Firstly, the gray distribution information and related information of the image were used to fuse redundancy, competition and complementarity of the image effectively, at the same time, the two-dimensional and three-dimensional observation space were constructed. The fast recursive algorithm of OTSU-H criterion was proposed. Secondly, the objective function of the fast recursive algorithm was respectively used as the fitness function of the four group intelligent algorithms of Cuckoo Search (CS) algorithm, Firefly Algorithm (FA), Particle Swarm Optimization (PSO) algorithm and Genetic Algorithm (GA). Finally, the optimized OTSU-H was introduced into the PCNN model to acquire the number of iterations automatically. The experimental results show that, compared with the original OTSU, the maximum entropy criterion, the image segmentation algorithms based on graph theory segmentation, pixel clustering segmentation and candidate region semantic segmentation, the proposed algorithm has better image segmentation effect, reduces the computational complexity, saves the storage space of the computer, and has strong anti-noise ability. In addition, the proposed algorithm has a wide range of applications with the characteristics of less time consumption and not need training.