Seeing in 3-D Stereo - Remote Sensing Application - Completely Remote Sensing, GPS, and GPS Tutorial
Seeing in 3-D Stereo

Prior to the 1930s, most topographic maps were made exclusively from field surveys. With the advent of aerial photography, and specifically as aerial surveying produced stereo pairs, planimetric maps became possible and special instruments were developed that could draw contours. Ground surveying is still necessary to precisely locate calibration points within photo scenes.. To understand how we apply stereophotogrammetry to contouring requires mastery of the concepts behind stereographic viewing.

This style of viewing is as familiar as seeing. When we alternately blink our eyes, objects appear to move slightly sideways. This shift is even more dramatic when we hold an index finger about 30 cm (18 in) in front of our nose and perform the above eye activity. The finger seems to jump left and right. This apparition comes from the principle of parallax. Parallax is the apparent displacement of a viewed point or small object at a distance that results from a change in the point of observation. For a person, that change in the point of observation could simply be from one eye to the other at a fixed location or from relocating from one viewing spot to another. Our line of sight from each eye is not quite parallel to the line between our nose and the selected target, but converges from the eye pair, so that the left eye sees a bit of the left side of the target not seen by the right eye, while the right eye sees a bit of the right side, missed by the left eye. In this way our eyes send a signal to the brain which, on further processing, creates the impression of depth.

This same effect lies at the heart of stereo viewing of photo pairs, taken either from two lateral positions on the ground, or, in aerial photography, as successive photo pairs with about 50% overlap along a single flight line, or with similar sidelap between pairs from adjacent flight lines, with some of the scene in common within the pairs (see page 10-3 for further details). When we position the stereo pair properly left-right and then view them through a stereoscope, the eye-brain reaction is an impression of surface curvature or relief, as though we're looking down from a plane at the ground. A pocket stereoscope consists of two lenses that we can adjust along a slide bar to be as far apart as our eyes; these are placed in a raised mount (on collapsible legs) which is positioned about six inches above the central region of the stereo pair.

The sense of relief may be exaggerated relative to reality. The degree of vertical exaggeration (VE) depends on the base to height ratio (B/H), which depends on the scale of the photos. The scale, in turn, shows the actual horizontal ground distance (B) between any two equivalent points, identifiable in the two photos, and the height (H) of the camera, during the exposure of each photo in the pair. These points will, of course, not occupy the same position in the two photos because of the forward motion of the imaging platform. The vertical exaggeration also depends on the apparent height (h) of the viewer's eyes and the breadth (b) between the eye centers of the particular viewer. So VE = (B/H)(h/b). VE typically ranges between 1+ and 6+ for B/H ranges between about 0.2 and 1.2.

There are several stereo pairs on this page and on page 11-8. If you try to print out either page to get these images to cut apart and examine in stereo, the length of the page will require several printed pages. To avoid this, we have placed some representative stereo pairs on a separate page, without text, which will reduce the number of printout pages needed. Access this page for printing by pressing on the word STEREO in blue/purple; return to this page or 11-8 by clicking on the browser BACK button.

If you have a pocket stereoscope, you may see the stereo effect by placing it, with legs extended, against the image below as it appears on the screen. This viewing usually doesn't work for most observers, so, it is probably necessary to print the pair, cut them apart, and then view them with the stereoscope. You likely will have to move one or the other laterally until the area viewed in common fuses visually into the stereo effect. (A few people can actually get a stereo effect from the images on this page as presented on the screen. But that ability is too uncommon to rely on, so we suggest instead that you go the printout route.)

We are reprinting below the first pair on the STEREO page, placing them in juxtaposition, separated enough for stereo viewing. Cut them from the printout page and position them. Staring down, try to get the stereo effect. You will probably have to shift the left laterally closer to, or farther from the right until the separation is just enough to bring about the effect (commonly, you start to see stereo but the two images aren't quite fused, so that you need to shift experimentally until the images coincide; also, when fusion occurs some people will see a third partial virtual image in the center between the two real images). This pair shows mature topography in a dissected hilly terrain, You may see the slight differences in shape (and shadowing) of the same hills in the two photos that result from the changed viewing positions. ()

Stereoscopic pair of a mature topography in a dissected hilly terrain.

If you do not have a stereoscope, you still may be able to get the 3-D effect without one. Many people can see in stereo with their unaided eyes. To test the likelihood of doing this, contact your two index fingers tip to tip at eye level about 20 cm (12 in) in front of your nose. You may see not only them, but an illusory "sausage," about an inch long, consisting of the two finger tips, that appears to "float" between the two real fingers. Moving the fingers closer to or farther from your eyes causes the sausage to expand or shrink. This won't work if you focus directly on the fingers, but should happen if you focus on "infinity," that is, gaze at long distance, so as to focus well beyond the fingers (the "vacant stare"). If this doesn't occur for you, it means you either have some physical eye limitation, or you psychologically don't believe that you can do this. Try this natural stereo viewing either on the image pair above or the one below (print them out, if you need to).

Now practice with this second image, which you also cut out:

Stereo image pair for a terrain in which sedimentary rock units are inclined (dipping).

We have scanned a pair of photos of the Little Dome (Wind River Basin, Wyoming) anticline (upward arching fold) and reproduced them below at rather large size. If you have not done so yet, access them by clicking on STEREO. After printing and cutting them apart, put the left one A to the left of the right B (or bottom one) and check their stereo expression. It may show up just fine or it may look funny. The positioning relative to the sequence in the flight line must be proper for normal stereo to take place. Since there was no information available to the writer (NMS) as to whether the top photo was taken first or second, you will have to experiment with choosing which photo to place left of the other, switching until the normal effect leaps out at you. We are printing the left one (A) of this pair on this page to make sure you have the right photos. Hold one in place and move the other laterally until stereo appears. You may have to bend (curl) the paper of one to see the corresponding area in the other. Be sure to trim any borders to eliminate any white effect.

Stereoscopic image (A) of an anticline in Wyoming.

Stereo Pairs from Space

Astronauts took the first Visible-IR stereo-pair photos from space during several missions. These photos led to an argument during the first days of Landsat-1's mission as to whether the sensors should be modified to acquire stereo. Landsat was not designed to acquire in stereo, because its downtrack images result from continuous scanning. However, its orbital track results in 10% to 40% (near the poles) sidelap coverage along adjacent orbits (previous or next day), which provides limited area stereo viewing, if cloud conditions are favorable. There is a Landsat stereo pair on the linked stereo page ([STEREO ]; if you haven't done so, print out this page and cut the Landsat images apart for appropriate overlaying). They consist of two complete Landsat MSS Band 7 images taken on October 10 and 11, 1972, thus just one day apart during a time of cloudfree conditions during the continued presence of a pressure High following an earlier storm. Determine where the scenes are geographically, the approximate percentage of sidelap that produces stereo, and the amount of stereo relief (governed by the B/H ratio).

This works also with Landsat images taken along contiguous orbits at various times of year. Check the stereo strips covering part of the now familiar Blue Mountain and other ridges near Harrisburg, PA which you printed out earlier (if not, go to STEREO): the top image (MSS Band 5) was taken in February (defoliate; low sun angle emphasizes relief), whereas that on the bottom was obtained the previous October (foliage still on trees). If you succeed in getting a stereo image using a printout, note the very low vertical exaggeration (about 0.3) because of the high H value (918 km).

Landsat stereo image (A) of the Blue Mountain area just north of Harrisburg, Pennsylvania, taken in February
Landsat stereo image (B) of the Blue Mountain area just north of Harrisburg, Pennsylvania, taken in October.

Now, try for 3-D in this Landsat false color composite stereo pairing of images of Katmandu and the surrounding Mountains in central Nepal, taken on March 20/21, 1977. The Environmental Institute of Michigan (ERIM) prepared the mount.

False color Landsat stereo image pair of the Katmandu area of central Nepal, March 20/21 1977.

Additional stereo images from SPOT are shown on page 11-9, along with the means to print them out.

There are two more stereo pairs on the STEREO printout page. One pair is a color set from SPOT. It is prepositioned to achieve stereo viewing for some people (if not, you must cut and separate). The other consists of two adjacent Landsat images of parts of eastern Pennsylvania, some of New Jersey, and a bit of New York. Look carefully for the upper right part of the left scene that has a common area in the left of the right scene (use the lakes as a guide). When separated, position them properly - in this case, you will need to shift one of the pair vertically since the two scenes are offset in their relative orbital locations. Try to get the common areas to fuse stereoscopically. If you do, you will notice tht the vertical relief is quite subdued: Owing to its high altitude, there is only a small B/H effect so that the stereo view shows little exaggeration (thus, Landsat has only limited value in this mode of viewing).

The next page continues the theme of stereo imagery acquired by satellites.