Medical Image Registration is a very important technology which enables doctors to perform better in
their daily tasks. Today doctors use different image modalities to give a better diagnosis. If a human experiences
lower back pain he/she will visit a doctor and the doctor will order an X-ray scan and a CT-scan. Because X-ray is
2D and CT is 3D it becomes difficult for doctors to compare between these two image modalities. In order to help
the doctors several image registering techniques have been developed which deform the CT into the desired pose taken
with the X-ray device. However, this process is still very slow, time consuming and inaccurate when we use a single X-ray
image to estimate the pose of the spine. Also the accurate result requires a special X-ray device which can take two images
at the same time from a different viewing angle. That means that the pose of vertebrae in the X-ray image will be the
same for both images no matter the viewing angle. This is very inefficient as the X-ray device is also expensive.
In this work I propose an algorithm on solving the depth error problem which occurs when a single X-ray image is used for pose estimation in the CT data. I compare my result to the pose estimation when two images of the same vertebrae pose are used. We believe that this gives a very accurate result. The solution is simple meaning that I propose that we can take multiple x-ray images and that they can be independent. Then we can estimate the pose for those single X-ray images and then interpolate the result to find the correct solution for the error. For this interpolation to work correctly the pose difference between two X-ray images should be very small and subtle. Also at least 3 X-ray images are needed and two of them must be from the same viewing angle. Of course the more images we have from the same view point the better our interpolation will be and this can serve as a training data set for future work