A sinusoidal density function, a rectangular density function and a set of circular and square features to test geometrical and contrast resolutions were used as input object functions. smooth object functions (both in geometry and in contrast/density) give rise to lower global errors compared to rough functions. Also the local error is lower in the regions where object function is smooth and large in the regions of high spatial frequency. The error is lower near the centre of reconstruction and higher towards the edges, for a smooth object function. It is demonstrated that for a smooth object function the error is proportional to the Laplacian, which manifests in edge enhancement.