Abstract: We consider human head as a rigid body, so the six parameters(3 of translation, 3 of rotation) that describe rigid body transformation can be applied to image registration. At first, the six parameters are given an initial value respectively. Then, they are optimized iteratively. The most optimized parameters are determined when we get the sum of squared differences between source image(the image to be registered) and reference image. After the acusector sting fMRI images are processed by 3D registration, most motion artifact can be eliminated, and the effect is satisfied.