Efficient content-based image retrieval of biomedical images is a challenging problem of growing research interest. This paper describes how X-ray images of the spinal columns are analyzed in order to extract vertebra regions and contours. Our goal is to develop a computer vision tool able to determine a global polygonal region for each vertebra in first time. After this step, we apply a polar signature system in order to extract the effective contour of each vertebra. Finally, we use an edge closing method exploiting a polynomial fitting. The aim is to propose a closed contours detection representing each vertebra separately. We suggest an application of the proposed method which consists on an evaluation of vertebra motion induced by their movement between two or several positions.