A min-cost-max-flow based algorithm for reconstructing binary image from two projections using similar images

The aim of this paper is to study the reconstruction of binary images from two projections using a priori images that are similar to the unknown image. Reconstruction of images from a few projections is preferred to reduce radiation hazards. It is well known that the problem of reconstructing images from a few projections is ill-posed. To handle the ill-posedness of the problem, a priori information such as convexity, connectivity and periodicity are used to limit the number of possible solutions. We use a priori images that are similar to the unknown image, to reduce the class of images having the same two projections. The a priori similar images may be obtained in many ways such as by considering images of neighboring slices or images of the same slice, taken in previous time instances. In this paper, we give a polynomial time algorithm to reconstruct binary image from two projections such that the reconstructed image is optimally close to the a priori similar images. We obtain a solution to our problem by reducing our problem to min cost integral max flow problem. © 2008 Springer-Verlag Berlin Heidelberg.