On the physical background of solid annealing process's properties, Simulated Annealing Algorithm is one sort of global optimization algorithm. This algorithm could solve the nonlinear problem and "Cycle Skipping" phenomenon of residual correction, but the disadvantages of redundant iteration and slow convergence speed also exist. Aiming at the above disadvantages, this paper proposed the following improved program: In the process of iteration, in order to decrease iterations and increase the convergence speed, the improved algorithm adopts new program to control the attenuation of temperature; The improved algorithm adopts new target function to increase the processing speed. By limiting the model disturbance in the condition of low temperature, the improved algorithm could obtain the optimal solution more quick. By processing the simulation data and actual seismic data of Algeria, the improved could decrease iteration times and improve convergence speed and efficiency effectively and enhance the continuity of reflection event, improve stacked sections' resolution and SNR (Signal to Noise Ratio).