The introduction of multi-slice helical computed tomography has fundamentally changed the way radiologists view CT images. Increasing numbers of clinical cases are examined with advanced visualization tools, such as maximum intensity projection, multi-planar reformation and volume rendering. It has been discovered that new image artefacts, which do not appear in the traditional two-dimensional reconstructed images, become visible in images generated by these new tools. In this paper, we investigate the causes of one such artefact, the Venetian blind artefact, which appears as bright-and-dark bands superimposed on three-dimensional images. We demonstrate that such an artefact is caused by the periodical noise variation in the reconstructed images. The image noise variation is, in turn, caused by the interaction of the noise inhomogeneity in the projections with the helical weights. An analytical formula is developed that accurately predicts the presence of such artefacts. Based on our analysis, several approaches are proposed for the artefact reduction or elimination.