From the algebraic point of view, image reconstruction is an under-determined problem, due to the fact that the projection measurements are a lot fewer than the unknown pixels. Discrete algebraic reconstruction technique (DART) is an algorithm used to cope with this situation. DART solves a discrete linear inverse problem by combining an algebraic reconstruction procedure with a threshold segmentation. In order to improve accuracy, the TvMin+DART algorithm modified the subroutines of DART by exploiting total variation minimization technique (TvMin). However, this algorithm was developed for only binary images. In this paper, our TvMin+DART algorithm will be generalized to handle multilabel images as well, and an experimental research to compare the proposed method with the original DART and the conventional filtered backprojection (FBP) will be presented.