Since there is deficiency in parameter estimation and classification of traditional change detection approach of remote sensing image based on Bayes decision rules. This paper adopts the method of dynamic updating the parameters and gains the posteriori probability function through these parameters. Then, Compute the criterion function through transforming the posteriori probability function to fuzzy subordinate function. Finally, Estimate the pixels that haven't been classified to extract the changed areas from remote sensing images. Through experiments, it proves that the change detection approach of remote sensing images based on fuzzy Bayes decision rules has a higher precision comparing to traditional change detection approach of remote sensing image based on Bayes decision rules and ERDAS.