In this study, similarity rates of the liver images which are obtained from different peoples are determined using 3D geometric transformation methods. The similarity is evaluated based on the numerical comparisons and visual results. 10 intact liver images which are drawn by the radiologists are used. Three geometric transformation methods scaling, rotating, and translating are consecutively applied to the liver images. All images are used both as atlas and as test images. The Dice coefficient values are calculated to show the similarity of each test image to atlas. The scaling, rotating, and translating amounts of the image are retained for the atlas which the similarity rate is highest. The liver images of different persons are similar to each other at an average rate of 67 -/+ 0.09 % according to Dice coefficient values which express the similarity. This study is presented as a step to prepare atlas database for segmentation of the injured liver.