Based on correlated imaging with pseudo-thermal light, Wu and Meyers independently experimentally showed that both positive and negative images can be obtained by using a novel algorithm. With the theory of statistical optics, we construct the model of thermal light, and demonstrate the novel algorithm for image reconstruction. To deal with the data attained in the correlated imaging, we reorder the intensity recorded by the bucket detector according to the value of fluctuation. For a given fluctuation range with all the records above or below a specific value, we can obtain either positive or negative images by calculate the correlation between the selected records of the bucket detector and the reference detector. Nevertheless, without correlated calculations, positive or negative images can be also produced by directly averaging the corresponding records of the reference detector with positive or negative fluctuations. Meanwhile, the visibility of imaging is greatly enhanced. This correspondence imaging method further demonstrates the importance of intensity fluctuations in the nonlocal imaging with thermal light. We also experimentally show the images that obtained by the correlation method and the positive-negative correspondence imaging method, respectively. Our results indicate that this novel algorithm has a better visibility than that of the conventional correlated imaging.