Development of 3-D shapes estimation by using single X-ray image

Abstract

X-ray images are heavily affected by noise which makes normal image processing not workable. This paper suggested a new method to identify the primary 3-D shape of an embedded object and its pose by using only single X-ray image. The image feature consists of corner points and edge/intersection lines of adjacent surfaces. The intensity of an X-ray image is attenuated exponentially with increasing the penetration thickness. The main finding is to model a precise exponential relationship to fit the variation of X-ray image intensity. It applied a least-square-method to the X-ray projection image and effectively extracted edges and intersection lines from the noise of X-ray image.