In wireless communication, distortions are induced into an image through both lossy source coding and an error prone channel. The combination thereof causes complex distortion patterns in the visual content which vary strongly in their spatial distribution, some being uniformly spread and others being locally clustered. In the context of image quality, it is of interest to identify as to how the distribution of the distortions impact on the visual attention of human observers and thus on the perceived quality. For this reason, we conducted an experiment in which fifteen participants were asked to rate the quality of eighty distorted images and their corresponding reference images, while we recorded their gaze patterns using an eye tracker. The test images contained a wide range of artifact magnitudes, both locally and globally distributed, and were created using a wireless channel simulation model. We then correlate the visual gaze patterns of the human observers with local energy measures of the distortions in order to identify the effect of distortion distributions on the viewing behaviour of human observers when judging image quality. The outcomes serve to improve the prediction performance of objective image quality metrics that we previously proposed.