Point cloud reconstruction with sub-pixel accuracy by slice-adaptive thresholding of X-ray computed tomography images

We present a methodology for the calculation of highly accurate point clouds from an image stack obtained with a two-dimensional industrial X-ray computed tomography scanner. The basic idea is to apply a different threshold value to each computed tomography image, which is then used for the calculation of the point cloud. A dense and smooth cloud can be obtained by interpolating the grey values between pixles. Our method combines both. An easy to implement thresholding algorithm is used for the estimation of a slice-by-slice threshold and an interpolating algorithm to reconstruct highly accurate point clouds.It is shown that the point clouds approximate the true dimensions of a scanned part with sub-pixel accuracy.