When you watch MAGIC EYE pictures you can recognize behind (or in front) of the screen 3D objects, which are encoded in the pictures or which can directly be seen. In order to learn the technique of 3D viewing you should start with pictures where the 3D objects are not encoded (see the aquarium). All the other pictures are with encoded 3D objects and it is more difficult to recocnize the 3D objects in those pictures.
This is how 3D viewing works: 3D pictures are almost periodical in horizontal direction. The point of the intersection of the line 'left eye - point of the picture' with the line 'right eye - point of the picture of the next period' is the appropriate point of the 3D object. That means 3D viewing works only with two eyes. A straight line through the left eye and the right eye must be parallel to the horizontal direction of the picture. Because the 3D picture is not exactly periodical it is possible to see structures in 3D. If the period has a high value the 3D object is faraway, if the period has a small value the 3D object is close to the viewer.

In order to get the best results in 3D viewing note this: And here a little bit more precise for the friends of mathematics:
                  | z-axis
           eye 1  |   eye 2
            O     |    O
            *     |   *
            *     |  *
            *     | *
            *     |*
            *     |
            *    *|
            *   * |    3D picture
            * *   |     (screen)     x-axis
            X     |
        3D object |
If the position of eye 1 is (x1, 0, z1), the position of eye 2 is (x2, 0, z2) = (-x1, 0, z1) and the position of the 3D object is (x3, y3, z3) then the appropriate positions in the 3D picture are (x1', y1', z1') and (x2', y2', z2'), respectively:
x1'=(z1x3-z3x1)/(z1-z3), x2'=(z1x3+z3x1)/(z1-z3), y1'=y2'=(z1y3)/(z1-z3) and z1'=z2'=0.

And here is a little program (Turbo Pascal source code) about that.

(c) Lutz Tautenhahn 1/99

go to the pictures