A novel Graphics Processing Units （GPU） accelerated level set model which organically combines the global fitting energy and the local fitting energy from different models and the weighting coefficient of the global fitting term can be adaptively adjusted， is proposed to image segmentation. The proposed model can efficiently segment images with intensity inhomogeneity regardless of where the initial contour lies in the image. In its numerical implementation， an efficient numerical scheme called Lattice Boltzmann Method （LBM） is used to break the restrictions on time step. In addition， the proposed LBM is implemented by using a NVIDIA GPU to fully utilize the characteristics of LBM method with high parallelism. The extensive and promising experimental results from synthetic and real images demonstrate the effectiveness and efficiency of the proposed method.In addition， the factors that can have a key impact on segmentation performance are also analyzed in depth.