Objective The virtual colonoscopy is a method that looking through the 3D structure of colon which reconstructed by CT or MRI images and detect colon tissue especially the polypus. Colon registration can greatly improve the efficiency and accuracy of polyp detection in virtual colonoscopy, but the traditional registration can’t accomplish that due to the drastic change in position of the patient in supine and prone. Method In this paper, we propose a new method of colon registration that can complete the image registration acquired in different position. First, extract the haustral fold that can reflect the information of fold structure. Get the matched haustral folds using the method of template matching and feature matching. Then do the non-rigid registration based in the center of matched haustral folds. Using b-spline registration to completed the fine registration of the two pictures finally. This method can retrieve the drastic change in the colon image, make the change of the two image within certain limits, and the traditional registration method can be used. Result There are about fifty percent haustral fold can be matched among ten sets of data. The rate of matching error is six percent and eight percent of haustral fold can not be find. The colonic deformation decreased after the rough registration of the folds.,that make the two colon images can be registered using the tradition register method. we had accomplished the registration of colon images finally. Conclusion When doing the rough registration of the two colon image, the deformation of two colon images can be corrected. However, when doing registration evaluation,we can not complete the assessment simply using the difference of gray-scale, so we should make some other criteria to assist the assessment.