Regional Studies: Use of Mosaics - Remote Sensing Application - Completely Remote Sensing tutorial, GPS, and GIS -
Regional Studies: Use of Mosaics

How Mosaics are Made

The notion of getting a much larger picture of a region by pasting images of individual scenes into a single composite goes back to the early days of aerial photography. The resultant product is called a mosaic. We construct mosaics a bit like jigsaw puzzles, except that we usually know the position of each air photo in advance. During an aerial mission (see page 10-1, which discusses aerial photography) we predetermine the flight lines for the airplane to follow, usually a back and forth pattern, much as a farmer plows a field.

The plane takes the photographs in sequence, such that there is always about 50% overlap (common area) between each successive picture, normally accomplished by an automated camera shutter at a timing interval controlled by altitude, air speed, and camera properties. On the next parallel line the pilot attempts to traverse the ground at a lateral distance that produces up to 40% sidelap. Both end-to-end and side overlaps allow for obtaining a three-dimensional or stereo effect (see page 11-3 for stereoscopic procedures and examples of 3-D views). However, the resulting pictures almost always display distortions, which comes from inexact navigation and aircraft wobble (in pitch, roll, and yaw) caused by turbulence.. There is also a notable distortion in an image outward from its center both because the camera is supposed to look vertically downward, but at times it’s canted off center, and the geometry of displacement imparted to objects towards the outside of a photo related to the increasing slant distances from center to edge. This effect decreases with higher altitudes (thus is lessened in space images) and we can also adjust the focal length and other camera geometry to diminish it. Tonal brightness also can be discernibly lower from the interior of the photo towards its boundaries.

In practice, we construct mosaics from the central parts of the component photographs, from which we trim enough to remove the more distorted, overlapping parts. Thus, they approximate rectified orthophotos. Large photo-mosaics have an added problem: the crews frequently fly the assigned flight lines on different days (or, less commonly, at different times of the same day), so that lighting and weather conditions often are not uniform. This problem is especially severe if the overflights happen weeks apart, in which case vegetation changes, along with the sun orientation due to seasonal shifts.. Photographic processing can compensate for some of these deviations, but uncontrolled (minimal adjustments) mosaics are typically rather patchy.

To retain maximum resolution, we must put together a photo-mosaic from individual scenes that we have not reduced in size. Consider making a mosaic from aerial photos having a scale of 1:62,500 (scale is discussed on page 10-1). Each photo, approximately 30 cm (12 in) on a side, covers almost 2.6 sq km (1 sq mi) of ground surface. At this scale, to create a mosaic representing ground dimensions of 6,400 sq km (2,500 sq mi), equivalent to 80 km (50 mi) on a side, we would need a "billboard" 15 m (50 ft) wide and high, requiring 2,500 photos, neglecting trimming requirements. This is impractical, so we almost invariably reduce mosaics in size and hence in scale and therefore, they have a notably lower resolution.

The scene below is a solid example of a typical uncontrolled mosaic. This is a series of high-altitude aerial photos (each about 18 km [12 mi] on a side) taken by NASA's U-2 aircraft along seven flight lines during late spring of 1972 in support of the writer's study of the geology of central Wyoming.

A uncontrolled mosaic (edge tones and distortions not corrected) made from a series of black and white aerial photos taken at high altitude from a NASA U-2 airplane; area shown in the Wind River Basin of central Wyoming.

Compare this scene, centered on the Wind River Basin, with this Landsat 1 Band 6 MSS scene that includes, and extends beyond, the mosaic. This was the prime image taken by the writer (NMS) into the field in central Wyoming just over a month after launch of ERTS-2 (to my knowledge I was the first to take any ERTS product to the area it covered so as to check on its usability):

Band 6 ERTS image of the Wind River Basin, obtained in early August of 1973; the version shown here was not adequately processed to achieve the best tonal balance.

To match the image with the U-2 mosaic, look for the Boysen Reservoir and Ocean Lake (round) that stand out in the MSS view but are hard to see in the mosaic. In that mosaic, the Owl Creek Mountains, with partial snow cover, lie along the top; the Sweetwater River is at the bottom; a cloud bank appears in left center. When examined full size on a light table, this black and white scene shows more ground detail, even in that tonal mode, than does the Landsat full image but the even-toned nature of the latter compensates somewhat for the information quality.

Shortly after returning from the field in 1973, the writer (NMS) arranged with a support contractor, GE Space Sciences Lab in Beltsville, MD, to reprocess black and white images of western and central Wyoming and part of adjacent Utah to optimize contrast and then to match tonal levels at joins so as to produce an ERTS mosaic - possibly the first ever made with this quality of space imagery. The result was both pleasing to the eye and informative. Here is that product:

Black and white mosaic of the western half of Wyoming and eastern Idaho.

It's hard to see any join lines in this mosaic. Main reason: the lack of vignetting, i.e., tonal gradation to the edges, because of the high altitude of Landsat relative to its image base size means that the outer parts have nearly the same tonal balance as the inner.

To the writer's knowledge, the next image is the first color mosaic of a State in the U.S. - Wyoming - ever made, again at the GE facility. This was used as a base for data plotting in the Wyoming project he conducted with members of the Geology Department of the University of Wyoming:

Uncontrolled Mosaic of almost all of the State of Wyoming.

As more Landsat images were acquired, and GE honed its skills in processing them, a new mosaic of Wyoming and parts of Utah and Montana was produced:

A later mosaic of Wyoming.