Three-dimensional reconstruction from truncated two-dimensional projections cannot be solved analytically without prior knowledge of the sample or access to the non-truncated projections. To suppress reconstruction artifacts in region of interest (ROI) or local tomography, an iterative algorithm has been devised based on back-projection and re-projection, assuming an approximately cylindrical shape of the entire sample of general homogeneity, which is especially applicable for micro-tomography of biopsy samples from biological tissue. Tomographic reconstruction is iteratively refined by minimizing the mismatch between an empty ROI and the reconstruction from the difference between measured sinogram and forward projected ROI reconstruction. By numerical simulation and experimental demonstration, it is shown that the algorithm is not only able to reconstruct quantitative gray values, but also to reduce artifacts of peripheral glow, and may lead to increased image sharpness. The method is particularly suitable for examination of biopsy/autopsy-samples of soft tissue by micro/nano-computed tomography.