However, another technique tested with another ATM flight showed more promise.
Images from a third spacecraft camera (An: looking straight down) were added. Shown below are Cf, An, and Ca images of the same site, with ATM values in color.
The ATM roughness data and image pixels under the flight path from all 3 cameras were then assembled into a 4-dimensional data model depicting the image values of each pixel along the X, Y, and Z axes of the model, and the logarithm of the ATM roughness values as color. This model showed a consistent grouping of similar roughness values (colors) in different parts of the image space.
With this data model, a simple procedure could be used to estimate the roughness values of other pixels in the image from their Cf, An, and Ca values, by performing a weighted average of the roughness values of the 4 closest model points in image space to the new pixel, using Gaussian weights. Only pixels within a cutoff distance of 0.1 image space units from the data model were estimated. Shown below is the result for a roughness of 5.63 cm (blue) to 972.07 cm (red). Black pixels were not estimated, as they were > 0.1 unit from any model points (e.g. there were no ATM data acquired over water). The multi-colored wavy features at bottom are clouds, and should be masked out. The data model used was extended with additional ATM data over the rocky parts of the image (darker pixels) and over the glacial outlet, to extend the range of input pixels estimated.
Shown next are plots of the distribution of image pixels (gray) overlaid with data model points (color, at left), and then overlaid with the roughness values of those pixels which were estimated (at right). About 75% of all pixels were estimated.
There was a good agreement between the measured roughness values (blue), and the estimated values (red):
To collect additional data for a better estimation model, all ATM flights (7) over Greenland during May 2007 were surveyed, and 21 sites/orbits for which there were good corresponding MISR image data in all 3 cameras were selected.
More on next page...
©Sky Coyote 2009