Reconstruction techniques, belonging to two different categories (series expansion and transform methods), are compared with regard to their applicability to line projections of a model density. It is demonstrated that in the case of this model the algebraic technique gives much better results than the filtered back projection (FBP) algorithm. However, it could not be a case for real data where a region of densities to be considered is very large comparing to the modelled one. In this connection we propose a new way of dealing with two-dimensional (2D) angular correlation of annihilation radiation (ACAR) data. Such a treatment, proposed for all reconstruction techniques, allows to improve reconstructed densities and filters the experimental noise.