In this paper, we propose an image quality measure that closely resembles the image quality measure used by the human visual system. In our application, we are more concerned with localized errors due to transmission than the global errors due to compression, and we develop an approach that is appropriate for this purpose. We use psychophysical experiments to determine the visibility threshold level for each frequency component of each image block. The visibility threshold levels are adjusted according to the background variation and intensity of the image block. Only those frequency components of an error that exceed the corresponding threshold level are included in the computation of the proposed image quality measure. The proposed image quality measure gives the PSNR of the errors that are visible to the eye and the percentage area of the extent of the visible errors in the image.