The psychology of perception is complex. I mean, when we look at a JPEG of a friend we see their image and not overlapping boxes of color variance -- this is because the brain forgives minor unexpected variation in chaotic regions. One form of steganography encrypts information by hiding it an image, but this only works in those noisy, chaotic parts of an image. You couldn't, for example, as easily hide information in an 8 bit image of a constant, medium grey area -- our brains would pick it right out, the same way they identify JPEG noise in poorly compressed text.
Or, to put it another way, you can certainly hide data in images. But you can't hide any data in any image from every observer. Most steganographic algorithms correct for this automatically.
