[lukifer]

Part of the difference is that audio is temporal, while an image is static. The odds of noticing an artifact in an well-compressed MP3 are small; the psycho-acoustic model is built on the assumption that many frequencies will be discarded by the brain anyway, and so the listener won't notice their absence.

However, an image doesn't change moment to moment, so the eye is free to wander and immerse, and thereby notice some oddly pixelated macroblocks that would have been ignored with a momentary glance.