A virtual image display apparatus comprises a pair of optical projection systems, each having a display device, a light source and a projection lens. The images formed on the display device are projected to an image location optical system, which comprises a light collecting lens and an optical diffusion device. The light rays from the optical projection systems are reflected at a translucent reflecting means, so that the light rays are respectively collected at right and left eyes of an observing person. As a result, the images can be recognized by the person as a virtual image in the distance. Since the optical diffusion device diffuses the light rays, in such a way that the light rays to one of observing eyes may not reach the other observing eye, so that the image information of the respective light rays can be recognized by the observing person, without causing a cross-talk.