How to extract useful information from noised image is always an important issue for image processing. Many methods have been proposed in image enhancement field. However, in these methods the noise is usually considered as harmful and should be removed as much as possible. Stochastic resonance is a very different method, in which the noise is regarded as a driver to push the stochastic resonance system to output enhanced image. In this paper, the cumulative gain is introduced and the sequence average is used to enhance the original image information which hidden in a noised image sequence produced by bistable stochastic resonance. We present the one-dimensional and two-dimensional stochastic resonance methods and discuss their performance in this paper. Experiments illustrate that the one-dimensional average stochastic resonance has the best performance considering the indicator PSNR and SSIM. Compared with traditional filters such as median and Wiener filters, the proposed methods have significant advantages.