Provided is a PRCA-based method for dynamically reestablishing a PET image. The method comprises: (1) performing collection and correction to obtain a coincidence counting matrix; (2) performing correction restriction on a PET measurement equation; (3) performing iterative estimation on PET dynamic concentration data; and (4) reestablishing each frame of PET image. In the present invention, different frames of collected data are regarded as a whole for reestablishment, and the relevancy of the PET data in time is fully used, so that the obtained result can embody the characteristic that the dynamic PET can show the time change of a target region; and secondly, in the present invention, a background and target region method is used, the interference to the target region caused by the background is reduced, in addition, the time and space correction are added in the reestablishment, the accuracy of the reestablishment result is higher, and the contrast between the target region and the background is improved. So that the reestablishment effect is more excellent than that of the traditional FBP and ML-EM algorithm, thereby providing higher medical value.