In this paper, an improved variable neighborhood placing based particle swarm optimization (VNP-PSO) algorithm, which is cooperative and population-based global searching swarm intelligence mataheuristics, is proposed. Generally, PSO has premature convergence property, which restricts its applications. Concerning this in mind, the proposed VNP-PSO may lead the particles, which are determined as trapped into local optimum, out from their local minima to the global optimum. To test the performances of VNP-PSO, experiments on four benchmark functions are conducted using the basic PSO and improved VNP-PSO respectively. Moreover, simulation on the optimization performances of VNP-PSO registering kidney ultrasonic image samples is conducted. The experiment results demonstrate that the proposed VNP-PSO method is an accurate, efficient optimization method, and suitable for medical image registration.