For optimization of the matching light source, this paper proposed a novel information recovery method based on the theory of stochastic resonance. This technique takes the advantage of noise to enlarge the dynamic range of image sensor when specifically added noise synchronizes with the image sensor system to activate the phenomenon of stochastic resonance. It investigated different types and strengths of noises to show the effect of variant distributions on the quality of retrieved information. The simulation results show that different types of noises have a corresponding strength, which makes the separability of output information reach extreme values. The results of locally overexposed experiment are consistent with the theoretic analysis, which also indicate that the recovery method of 2D bar code based on stochastic resonance could suppress the high light and improve image contrast.