Optical techniques such as speckle photography and rainbow schlieren deflectometry yield path integrated measurements of deflection angle instead of the local field variable such as density, temperature and/or species concentration. Thus, a reconstruction algorithm is employed to obtain the local properties from the path-integrated measurements. Cross-beam correlation (CBC) algorithm provides the link between path-integrated and local field statistics in time-averaged axisymmetric turbulent flows. Path-integrated statistics are obtained using orthogonal light rays crossing within the turbulent flow. The algorithm assumes local isotropy and negligible correlation between points on orthogonal beams, which is valid strictly in fully turbulent flows. In this study, noise-free synthetic scalar turbulence data are generated and used to determine how different assumptions in the CBC algorithm affect the reconstruction accuracy. Results show that the reconstruction accuracy is excellent for a narrow correlation function (or small integral length scale), while significant errors are incurred in case of a wide correlation function. A procedure to improve the reconstruction accuracy of the CBC algorithm is proposed in this study. © Springer Science+Business Media B.V. 2009.