Abstract: Based on Markov random field (MRF) image segmentation frame, this paper presents a better partitioning algorithm for surface mount technology (SMT) solder segmentation which is the simulated annealing algorithm based on Gibbs sampling, and discusses the main factors which affect the segmentation effect. Finally this algorithm is compared with the traditional Gibbs algorithm and simulated annealing algorithm. It is shown from the experiment that the proposed algorithm only needs a few artificial interactions to reduce the sample dimensions and decrease the optimization and convergence time, and it has the fastest convergence speed to obtain the global optimization, with high segmentation success rate and accurate result. The segmentation result provides the guarantee for the further quality analysis of solder joints.