A Method for Feature Detection in Binary Tomography

Wagner Fortes, Kees Joost Batenburg

Abstract: Binary tomography deals with the problem of reconstructing a binary image from its projections. Depending on properties of the unkown original image, the constraint that the image is binary enables accurate reconstructions from a relatively small number of projection angles. Even in cases when insufficient information is available to compute an accurate reconstruction of the complete image, it may still be possible to determine certain features of it, such as straight boundaries, or homogeneous regions. In this paper, we present a computational technique for discovering the possible presence of such features in the unknown original image. We present numerical experiments, showing that it is often possible to accurately identify the presence of certain features, even without a full reconstruction.