To solve the problems that the fusion rule in available image fusion methods can not be adjusted adaptively according to successive application, and the advantages of different fusion methods can not be integrated, a new image fusion method is proposed based on particle swarm optimization. Firstly, the DBSS (2, 2) discrete wavelet transform and gradient pyramid fused images are used to create images as initial particles. Then, according to the requirements of successive application, the object function is constructed by the weighted sum of several image quality assessment criteria. Finally, according to object function, the fused image is obtained by particle swarm optimization. Two sets of images are assessed from visual inspection and quantitative evaluation (standard deviation, average gradient, entropy, spatial frequency, correlation coefficient and mean root cross entropy) to validate the feasibility of proposed method.